Identification and Structure–Activity Relationship Studies of Small-Molecule Inhibitors of the Methyllysine Reader Protein Spindlin1

Article abstract of DOI:10.1002/cmdc.201600362

The methyllysine reader protein Spindlin1 has been implicated in the tumorigenesis of several types of cancer and may be an attractive novel therapeutic target. Small-molecule inhibitors of Spindlin1 should be valuable as chemical probes as well as potential new therapeutics. We applied an iterative virtual screening campaign, encompassing structure- and ligand-based approaches, to identify potential Spindlin1 inhibitors from databases of commercially available compounds. Our in silico studies coupled with in vitro testing were successful in identifying novel Spindlin1 inhibitors. Several 4-aminoquinazoline and quinazolinethione derivatives were among the active hit compounds, which indicated that these scaffolds represent promising lead structures for the development of Spindlin1 inhibitors. Subsequent lead optimization studies were hence carried out, and numerous derivatives of both lead scaffolds were synthesized. This resulted in the discovery of novel inhibitors of Spindlin1 and helped explore the structure–activity relationships of these inhibitor series.

Full text of DOI:10.1002/cmdc.201600362

DOI: 10.1002/cmdc.201600362
Full Papers
Identification and Structure–Activity Relationship Studies
of Small-Molecule Inhibitors of the Methyllysine Reader
Protein Spindlin1
Dina Robaa+,[a] Tobias Wagner+,[b] Chiara Luise,[a] Luca Carlino,[a] Joel McMillan,[c, d] Ralf Flaig,[d]
Roland Schꢀle,[c, e, f] Manfred Jung,[b, e, f] and Wolfgang Sippl*[a]
The methyllysine reader protein Spindlin1 has been implicated
in the tumorigenesis of several types of cancer and may be an
attractive novel therapeutic target. Small-molecule inhibitors of
Spindlin1 should be valuable as chemical probes as well as po-
tential new therapeutics. We applied an iterative virtual screen-
ing campaign, encompassing structure- and ligand-based ap-
proaches, to identify potential Spindlin1 inhibitors from data-
bases of commercially available compounds. Our in silico stud-
ies coupled with in vitro testing were successful in identifying
novel Spindlin1 inhibitors. Several 4-aminoquinazoline and qui-
nazolinethione derivatives were among the active hit com-
pounds, which indicated that these scaffolds represent promis-
ing lead structures for the development of Spindlin1 inhibitors.
Subsequent lead optimization studies were hence carried out,
and numerous derivatives of both lead scaffolds were synthe-
sized. This resulted in the discovery of novel inhibitors of
Spindlin1 and helped explore the structure–activity relation-
ships of these inhibitor series.
Introduction
Epigenetic mechanisms play a central role in the control of
gene expression and are thus pivotal for regulating vital pro-
cesses like development, cellular differentiation and prolifera-
tion.[1] These mechanisms are highly complex and dynamic in
nature and involve diverse processes, most importantly DNA
methylation, histone modification/remodeling and noncoding
RNAs.[2] Owing to the compelling evidence linking dysregula-
tion of the epigenetic machinery to a wide variety of diseases,
like cancer,[3] neurodegenerative disorders[4] and cardiovascular
diseases,[5] the use of epigenetic modulators represents
a highly promising therapeutic approach.
for the cell to modify the DNA-histone and histone-histone
binding interactions, thus allowing the chromatin to switch be-
tween transcriptionally active and inactive forms. This process
is governed by three groups of proteins which introduce,
remove and recognize specific PTMs, and are accordingly clas-
sified as writers, erasers and readers. To date, numerous types
of PTMs have been recognized, such as acetylation, methyla-
tion, phosphorylation, sumoylation and ubiquitinylation of his-
tone tails.[1] PTMs do not only control the DNA transcriptional
processes by directly altering the chromatin structure, but they
can also recruit specific ‘reader’ domains to guide further
downstream signaling cascades.[6] These reader proteins com-
prise structurally conserved domains, able to recognize specific
histone marks.[6a,7] In recent years, diverse reader domain fami-
lies have been discovered, including readers of methyllysine,
acetyllysine, methylarginine, as well as phosphorylated serine
and threonine histone marks.[6a,8] Indeed, emerging evidence
has clearly indicated that some reader proteins are implicated
in the genesis and progression of human diseases, most im-
portantly cancer. However, the exact biological role of most
reader proteins and their validity as novel therapeutic targets
has not yet been completely clarified. Small-molecule inhibitors
selectively disrupting the binding of histone code readers to
their endogenous substrates, should, thus, be highly valuable
as chemical probes and as potential therapeutics for the treat-
ment of various diseases. Quite recently, a breakthrough has
been achieved by the discovery of BET bromodomain inhibi-
tors, some of which have already reached clinical trials.[9] So
far, only few small-molecule inhibitors of methyllysine readers
have been disclosed, including inhibitors of the L3MBTL1 and
L3MBTL3,[10] the PHD finger of JARID1A[11] and PygoPHD,[12]
the CBX chormodomains CBX7 and CBX4,[13] and the Tudor do-
Histone modifications, also referred to as post-translational
modifications (PTMs) of histone tails, are a highly dynamic way
[a] Dr. D. Robaa,+ C. Luise, Dr. L. Carlino, Prof. Dr. W. Sippl
Institute of Pharmacy, Martin Luther University of Halle-Wittenberg, Wolf-
gang-Langenbeck-Str. 4, 06120 Halle/Saale (Germany)
E-mail: wolfgang.sippl@pharmazie.uni-halle.de
[b] Dr. T. Wagner,+ Prof. Dr. M. Jung
Institute of Pharmaceutical Sciences, University of Freiburg, 79104 Freiburg
(Germany)
[c] J. McMillan, Prof. Dr. R. Schꢀle
Department of Urology and Center for Clinical Research, University of Frei-
burg Medical Center, 79106 Freiburg (Germany)
[d] J. McMillan, Dr. R. Flaig
Diamond Light Source Ltd., Harwell Science & Innovation Campus, Didcot,
Oxfordshire OX11 0DE (UK)
[e] Prof. Dr. R. Schꢀle, Prof. Dr. M. Jung
German Cancer Research Center (DKFZ), Heidelberg (Germany)
[f] Prof. Dr. R. Schꢀle, Prof. Dr. M. Jung
German Cancer Consortium (DKTK), Freiburg (Germany)
[+] These authors contributed equally to this work.
Supporting information for this article can be found under http://
dx.doi.org/10.1002/cmdc.201600362.
ChemMedChem 2016, 11, 1 – 13
1
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
mains 53BP1[14] and Spindlin1.[15] Detailed reviews are available
in the literature.[16] Given the advanced state of development
of inhibitors of reversible histone methylation,[17] many poten-
tial inhibitors can be envisaged for methyl readers as well.
Spindlin1 is a methyllysine reader protein, which recognizes
the H3K4me3 (H3 trimethylated at lysine 4) mark. Spindlin1
was found to be overexpressed in several types of malignant
tumors, including ovarian cancer, certain types of liver carcino-
mas, non-small-cell lung cancers and liposarcoma.[18] Upregula-
tion of spindlin1 gene was shown to induce morphological
changes in cells, increased cellular proliferation, abnormal mi-
tosis and chromosomal instability.[19] Different studies have im-
plied possible mechanisms behind the role of spindlin1 in tu-
morigenesis. Spindlin1 was indicated to promote the activity
of the Wnt/TCF-4 signaling pathway, leading to enhanced
cancer cell proliferation.[18c] Moreover, the contribution of
spindlin1 to liposarcoma was attributed to an enhanced RET
signaling pathway as a result of an overexpression of GDNF.[18a]
The abovementioned findings clearly demonstrate that Spind-
lin1 may be an attractive therapeutic target. Recently, A366
(Figure 1), also a potent inhibitor of the protein lysine methyl-
transferase G9a,[20] was found in our laboratories by focused li-
brary screening using a modular assay platform to inhibit
Spindlin1.[15]
Spindlin1 is composed of three structurally homologous
Tudor-like domains, where the binding to H3K4me3 is chiefly
mediated by domain II.[21] As characteristic for all methyllysine
readers,[6a,22] the trimethylated lysine is embedded in an aro-
matic cage, which, in the case of Spindlin1, is composed of the
aromatic amino acid residues Phe141, Trp151, Tyr170 and
Tyr177. The binding of the trimethylated lysine to the residues
of the aromatic cage is mainly mediated via cation–p interac-
tions and to a lesser extent hydrophobic van der Waals con-
tacts. As observed in the resolved crystal structures of Spind-
lin1 in complex with the histone substrate, the N-terminal resi-
dues (Ala1-Arg2-Thr3) of the histone peptide significantly con-
tribute to the binding (Figure 2). Ala1 and Arg2 of the histone
peptide are inserted in electronegative regions, where salt
bridges are formed between Ala1 and the protein residues
Glu142 and Asp189 and between Arg2 with Asp184. Mutations
of the aforementioned protein residues (Asp184 and Asp189)
led to decreased binding affinities, indicating the importance
of these interactions for the recognition and binding to the
histone peptide.[21] Moreover, the backbone of Thr3 of the pep-
tide undergoes two H-bond interactions with the backbone of
Glu142. Several water-mediated H-bond interactions can also
be observed.[21,23] Notably, the binding affinity of H3K4me3 is
enhanced when the H3R8 residue is asymmetrically dimethy-
lated (H3K4me3Rme2a). Here, the dimethylated arginine resi-
due of the histone peptide is inserted into the aromatic cage
of Tudor domain I.[23]
To search for novel Spindlin1 inhibitors, we primarily used
an iterative virtual screening (VS) campaign which combined
Figure 1. Structure of the Spindlin1 inhibitor A366. IC50 value taken from
Wagner et al.[15]
Figure 2. Crystal structure of Spindlin1 in complex with H3K4me3 (PDB ID: 4H75) shows A) three Tudor domains, where the peptide mainly binds to do-
main II. B) The N-terminal residues of the peptide (orange sticks) undergo several interactions with the binding pocket residues of Spindlin1 (green sticks).
Water molecules are shown as red spheres, and H-bond interactions as yellow dashed lines.
&
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
2
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
a variety of structure- and ligand-based approaches. Based on
the experimentally determined binding mode of the peptide
with domain II of Spindlin1, pharmacophore models were gen-
erated and subsequently used to screen databases of commer-
cially available compounds [ChemBridge Corp., San Diego, CA
92121 (USA) and ChemDiv Inc., San Diego, CA 92121 (USA)].
The selection of the most promising candidates for in vitro
testing was accomplished by docking the obtained pharmaco-
phore hits into the crystal structure of Spindlin1 (PDB ID:
4H75)[21] using the program Glide;[24] where only the top
ranked compounds showing significant interactions with the
residues of the binding site were considered.
The identification of several validated hits by means of an
established in vitro assay paved the way for further studies.
Here we used two strategies concomitantly. These entail
a ligand-based approach combined with docking studies in
order to find structural analogues of the obtained hits, and
secondly, lead optimization studies. In the latter, various deriva-
tives of the obtained hit compounds were synthesized in an
attempt to increase their inhibitory activity for Spindlin1 and
to draw conclusive SAR. The lead optimization step was con-
stantly guided by docking studies.
The inhibitory activity of the in silico hits as well as the syn-
thetized derivatives was determined by means of an AlphaLISA
assay. The commercially available TruHitsTM Kit was used to rule
out false positive hits emerging from the AlphaLISA assay. A
detailed description of the assays has been recently pub-
lished.[15]
Figure 3. Structure-based pharmacophore model generated using Ligand-
Scout. Blue spheres indicate a cationic feature (CAT), two green spheres and
an arrow denote a H-bond donor feature (HBD, vector type), and two red
spheres and an arrow denote a H-bond acceptor feature (HBA, vector type).
The N-terminal residues of H3K4me3 are shown as grey sticks, the binding
pocket residues as cyan sticks, and water molecules as red spheres.
Results and Discussion
Iterative virtual screening campaign
Pharmacophore- and docking-based screening using Ligand-
Scout and Glide
the first four N-terminal residues of the peptide. With the ex-
ception of the cationic feature denoting the positively charged
ammonium group of K4me3 (CAT-1, Figure 3), it is difficult to
estimate which other features should be considered as essen-
tial. Hence, based on the originally generated model, we pro-
ceeded by constructing six different pharmacophore models,
each containing three pharmacophoric features. All models in-
cluded the cationic feature describing the positively charged
ammonium group of K4me3 (CAT-1) while the other features
were alternatively combined. Two databases of commercially
available compounds, namely ChemBridge and ChemDiv, were
retrieved from the ZINC website (zinc.docking.org)[27] and
cured using MOE[28] modeling package. To decrease the
number of compounds for the subsequent pharmacophore
screening procedure, all compounds with a molecular weight
>500 as well as compounds containing no positively charged
N atoms (N+) or H-bond donors were filtered out. Pharmaco-
phore screening of the filtered databases resulted in a total of
3120 unique hits, which were subsequently docked using Glide
in Standard Precision (SP) mode. The docking results were ana-
lyzed and only the top-scored compounds showing a protonat-
ed aliphatic nitrogen rightly positioned in the aromatic cage
were considered. Following visual inspection and selection of
the compounds showing favorable interactions with the bind-
Crystal structures of Spindlin1 in complex with histone pep-
tides clearly illustrate that the main interactions are formed be-
tween the N-terminal residues (Ala1-Gln5) of the histone sub-
strate and residues of Tudor domain II.[21,23]
Our main strategy was to search for small-molecules which
could compete with the binding of the substrate to this
domain by mimicking the peptide interaction mode. We first
approached this by performing a pharmacophore-based VS fol-
lowed by docking studies in order to search for potential
Spindlin1 inhibitors in databases of commercially available
compounds. The pharmacophore-based approach serves as
a primary filter to select the compounds which could possibly
mimic the interactions of the histone peptide. Meanwhile,
docking should enable a better evaluation of the pharmaco-
phore hits by means of their docking scores as well as their
predicted binding interactions. To this end, the crystal structure
of Spindlin1 in complex with H3K4me3 peptide, PDB ID:
4H75,[25] was retrieved from the RCSB Protein Data Bank (PDB,
www.rcsb.org), and pharmacophoric features describing the
protein–peptide interactions were assigned using LigandSc-
out.[26] As can be seen in Figure 3, the resulting pharmaco-
phore contained seven pharmacophoric features generated for
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
3
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
ing site, a total of 71 compounds were purchased and subject-
ed to in vitro screening.
crystal structure of Spindlin1 (PDB ID: 4H75) using Glide in SP
mode, and a total of 23 substances were purchased and tested
using the established AlphaLISA assay (Table 2) and verified by
the TruHits assay.
Four compounds, namely the 4-aminoquinazolinethione de-
rivatives 1a and 1b as well as the bicyclic derivative 2a and
the N-(azepan-3-yl-propyl)benzamide 3a, showed significant
inhibition of the H3K4me3 binding to Spindlin1 in the AlphaLI-
SA assay (Table 1). The performed screening process is outlined
in Figure 4.
The described screening process is outlined in Figure 4. The
applied ligand-based VS was successful in finding four active
structural analogues of the 4-aminoquinazolinethione deriva-
tives, namely the thioquinazolinones (4a–d). A detailed discus-
sion of the SAR of the thioquinazolinone and 4-aminoquinazo-
linethione derivatives is given in the following section. Out of
the nine purchased analogues of 2a, only the structural ana-
logue 2e showed a significant inhibitory activity (IC50 =32 mm).
On the other hand, the two closely related homologs of 3a
(3b and 3c) demonstrated only a very weak disruption of the
binding of H3Kme3 to Spindlin1. Overall the performed VS
campaign was successful in identifying six micromolar inhibi-
tors of Spindlin1 (IC50 of 23–86 mm) representing four different
chemical scaffolds. These findings encouraged us to further
proceed with lead optimization studies focusing on the 4-ami-
noquinazolinethione and thioquinazolinone derivatives as lead
scaffolds. None of the VS hits were recognized as potential
PAINS.[29]
Table 1. Validated hits obtained from the pharmacophore-based VS.
Compd
Structure
IC50 [mm][a]
1a
33.2ꢀ4.0
1b
31% @ 50 mm
2a
3a
86.1ꢀ17.7
36.7ꢀ2.9
Structural modifications and SAR studies on the lead 4-ami-
noquinazolinethione and thioquinazolinone scaffolds
In an attempt to attain compounds with improved inhibitory
activities against Spindlin1, further derivatives of the lead scaf-
folds were synthesized. This should also serve to deduce the
SAR of these new series of Spindlin1 inhibitors.
[a] IC50 values are the meanꢀSEM of three experiments performed in
triplicate; percent inhibition values provided at the indicated concentra-
tion.
As depicted in Figure 5, both lead structures 1a
and 4c showed similar binding modes, where the ali-
phatic basic side chain acts as a lysine-mimetic and is
accommodated in the aromatic cage of domain II.
Additional H-bond interactions are formed between
the quinazoline scaffold and the backbone carbonyl
of Glu142 as well as with the side chain of Asp95, in
the case of the 4-aminoquinazolinethione derivative.
Meanwhile, the dimethoxy or diethoxy substituents
do not undergo any interactions with the binding
pocket residues. It is noteworthy that the electroneg-
ative surface groove of Spindlin1, containing Asp184,
Asp189 and Gln180, is not addressed by the ligands.
Therefore, we adopted a multifold strategy for struc-
tural derivatization of both lead scaffolds which in-
cluded the variation of the alkoxy groups at the 6-
Figure 4. Chart of the overall screening process. The pharmacophore-based screening is
depicted on the left side, while the subsequent ligand-based screening process is shown
on the right. TC=Tanimoto coefficient.
and 7-position of the quinazoline ring as well as probing differ-
ent lysine-mimetic moieties.
Ligand-based VS
To obtain further derivatives and/or structural analogues of the
active hits, a 2D-chemical similarity search for their close struc-
tural neighbors was carried out. Using MACCS structural key
fingerprints (MOE), the ChemBridge and ChemDiv libraries
were scanned, whereby a Tanimoto coefficient of 0.75 (for 1a)
or 0.8 (for 2a and 3a) was set a similarity metric. The per-
formed screening resulted in a total of 492, 1501 and 522
unique hits for related compounds of 1a, 2a and 3a, respec-
tively. The retrieved hits were consequently docked into the
Chemistry
Several derivatives of the 4-aminoquinozolinethione series
were synthesized, which contained either a methoxy or
benzyloxy substituent at position 7 of the heterocyclic ring
system in addition to different lysine-mimetic groups at posi-
tion 4. To this end, 3-hydroxy-4-methoxybenzonitrile (5) was
converted into the corresponding 3-benzyloxy derivative 6 by
&
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
4
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Table 2. Inhibitory activity of the hits obtained from the chemical similarity search in the Spindlin1 AlphaLISA assay.
Compd
Structure
IC50 [mm][a]
Compd
Structure
IC50 [mm][a]
71% @ 67 mm
(Q 49% @ 50 mm)
86% @ 100 mm
(Q 49% @ 100 mm)
1c
4e
1d
1e
18% @ 100 mm
4 f
50% @ 100 mm
98% @ 100 mm
(Q 74% @ 100 mm)
32% @ 100 mm
2% @ 25 mm
2b
3% @ 100 mm
0% @ 25 mm
1 f
18% @ 67 mm
31% @ 70 mm
2c
16% @ 100 mm
0% @ 25 mm
1g
2d
1h
23% @ 70 mm
2e
2 f
32.1ꢀ2.6
9% @ 100 mm
0% @ 25 mm
4a
4b
22.7ꢀ3.8
77% @ 100 mm
6% @ 25 mm
63% @ 100 mm
7% @ 25 mm
3b
3c
24% @ 100 mm
0% @ 25 mm
4c
25.0ꢀ5.8
85% @ 100 mm
61% @ 50 mm
33% @ 25 mm
4d
[a] IC50 values are the meanꢀSEM of three experiments performed in triplicate; percent inhibition values provided at the indicated concentration. Quench-
ing (Q) was assessed with the commercially available TruHits Kit.
using Williamson ether synthesis[30] using benzyl bromide. Sub-
sequent nitration using HNO3/H2SO4 in glacial acetic acid/acetic
anhydride yielded the 2-nitrobenzonitrile derivative 7, which
was then reduced by iron powder in ethanol/acetic acid/water
following a procedure reported by Gamble et al.[31] using ultra-
sound to yield the corresponding aromatic amine 8a.
commercially acquired or separately synthesized aliphatic
amines (see the Supporting Information) in the presence of
triethylamine (TEA) under microwave irradiation. The incorpo-
ration of the side chain at position 4 of the quinazoline scaf-
fold has previously been proposed to occur via a rearrange-
ment mechanism.[32] (Scheme 1)
The corresponding 2-amino-4,5-dimethoxybenzonitrile (8b)
was commercially acquired. Treatment of the aromatic amines
8a,b with thiophosgene yielded the corresponding isothiocya-
nates 9a,b, which were finally converted into the desired 4-
aminoquinazolinethione derivatives 1i–m by reacting with
Meanwhile, methyl 4-hydroxy-3-methoxybenzoate (10a)
served as the starting point for the desired 6-methoxy-7-benzy-
loxy-thioquinazolinone derivatives 4g–w. Williamson ether syn-
thesis using benzyl bromide, K2CO3 and microwave irradiation
afforded the corresponding methyl dialkoxybenzoates 11 a,
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
5
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
which were nitrated using HNO3 in glacial acetic acid to the
4,5-dialkoxy-2-nitrobenzoate derivatives 12a. The 2-nitroben-
zoate derivative bearing an N,N-dimethylaminopropyloxy sub-
stituent at position 7 (12b) was obtained by the alkylation of
methyl 6-methoxybenzoate with chloroiodopropane followed
by nitration and the conversion of the alkyl chloride to the cor-
responding amine using a Finkelstein reaction. Reduction of
the nitro group was afforded by iron powder and the obtained
aromatic amines 13d were reacted to the corresponding iso-
thiocyanates 14c using thiophosgene in a dichloromethane/
water mixture. Meanwhile the 6,7-dimethoxy-isothiocyanato-
benzoate 14a was directly purchased. Analogous to previously
reported procedures,[32–33] ring closure and the simultaneous
incorporation of an aliphatic, basic side chain (lysine-mimetic
group) at N3 of the 6,7-dialkoxythioquinazolinone ring was
achieved by reacting with the respective primary amine in the
presence of TEA under microwave irradiation (Scheme 2). The
6,7-unsubstituted thioquinazolinone derivatives 4g–m were
obtained directly from methyl 2-isothiocyanatobenzoate by
the reaction with the respective primary amine using the de-
scribed procedure.
The inhibitory activity of all newly synthesized thioquinazoli-
none and 4-aminoquinazolinethione derivatives against Spind-
lin1 was assessed by the established AlphaLISA assay and the
results were verified by the TruHits assay. Only compounds
showing significant inhibition at 100 mm concentration and no
significant quenching in the TruHits assay were further select-
ed for IC50 determination. (Tables 3 and 4, for the thioquinazoli-
none and 4-aminoquinazolinethione, respectively). Moreover,
preliminary selectivity studies were performed by testing some
selected compounds (1i, 1j, and 4o) for their inhibitory activi-
ty against the histone methyltransferases EZH2, both in a pep-
tide based and an oligonucleosome based assay, and Dot1L.[34]
No inhibition was observed in any assay at 80 mm concentra-
tion of the substances.
Altogether, a set of 13 4-aminoquinazolinethione derivatives
(1a–m, five newly synthesized and eight purchased) and 23
thioquinazolinones (4a–w, 18 newly synthesized and five pur-
chased) was acquired, which served to explore the SAR of
these new classes of Spindlin1 inhibitors.
Figure 5. Predicted binding modes for the active compounds 1a and 4c.
A) Predicted binding mode of 1a. B) Predicted binding mode of 4c. The
docking poses are shown in beige, Spindlin1 residues involved in the inter-
actions are depicted in green, the water molecule (HOH1416) as a red
sphere, and H-bond interactions as yellow dashed lines.
Concerning the 4-aminoquinazolinethione series 1a–m, a fo-
cused number of structural modifications was investigated, in-
Scheme 1. Reagents and conditions: a) benzyl bromide, TBAI, K2CO3, acetone, microwave; b) H2SO4/HNO3, HOAc/Ac2O, 08C!RT; c) Fe powder, ultrasound
bath, 308C; d) CSCl2, K2CO3, 30% H2O, CH2Cl2, 08C!RT; e) NH2-(CH2)n-NR’R’’, TEA, THF, microwave.
&
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
6
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Scheme 2. Reagents and conditions: a) benzyl bromide, TBAI, K2CO3, acetone, microwave; b) HNO3, HOAc, 08C!RT; c) Fe powder, ultrasound bath, 308C;
d) CSCl2, K2CO3, 30% H2O, CH2Cl2, 08C!RT; e) NH2-(CH2)n-NRR, TEA, THF, microwave; f) 1-chloro-3-iodopropane, K2CO3, MeCN, microwave: 90 W, 1108C,
30 min; g) NaI, TBAI, acetone, microwave: 40 W, 1008C, 10 min; h) dimethylamine, K2CO3, 60 W, 1208C, 20 min.
Table 3. Inhibitory activity of the synthesized 4-aminoquinazolinethione
Table 4. Inhibitory activity of the synthesized thioquinazolinone deriva-
derivatives in the Spindlin1 AlphaLISA assay.
tives in the Spindlin1 AlphaLISA assay.
Compd
Structure
IC50 [mm][a]
5.3ꢀ0.03
Compd
Structure
IC50 [mm][a]
4g
6% @ 100 mm
1i
4h
4i
12% @ 100 mm
0% @ 100 mm
61% @ 100 mm
1j
5.5ꢀ0.4
3.5ꢀ0.5
1k
1l
4j
98% @ 100 mm
83% @ 25 mm
(Q 62% @ 100 mm)
4k
4l
7% @ 100 mm
27% @ 100 mm
20% @ 25 mm
1m
9% @ 100 mm
11% @ 100 mm
[a] IC50 values are the meanꢀSEM of three experiments performed in
triplicate; percent inhibition values provided at the indicated concentra-
tion. Quenching (Q) was assessed with the commercially available TruHits
Kit.
4m
4n
4o
34% @ 100 mm
20% @ 25 mm
cluding the variation of the lysine-mimetic moiety at the 4-
amino group and the replacement of the 7-methoxy group by
a benzyloxy moiety. Our data clearly indicate that the aliphatic
basic substituent at the 4-amino group acts as a lysine mimic;
compounds bearing amide (1m) or aromatic groups as termi-
nal substituents (1c–e) showed hence very weak inhibitory ac-
20.7ꢀ1.8
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
7
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
Table 4. (Continued)
Compd
Structure
IC50 [mm][a]
16% @ 100 mm
0% @ 25 mm
4p
4q
4r
4s
4.5ꢀ0.5
5.7ꢀ0.3
7.3ꢀ1.0
56% @ 100 mm
29% @ 25 mm
4t
4u
34.5ꢀ3.8 mm
8% @ 100 mm
0% @ 25 mm
4v
30% @ 100 mm
0% @ 25 mm
4w
[a] IC50 values are the meanꢀSEM of three experiments performed in
triplicate; percent inhibition values provided at the indicated concentra-
tion.
tivities. Moreover, the nature of the terminal basic moiety
seems to have a significant effect on the activity. In the case of
the 6,7-dimethoxy derivatives, terminal N,N-diethylamino or
azepane moieties resulted in the highest inhibitory activity (1i:
IC50 =5.3 mm, and 1a: IC50 =33.2 mm). While the derivative bear-
ing a terminal pyrrolidine group still exhibited a considerable
inhibition of Spindlin1 (1b, IC50 >100 mm), the corresponding
morpholino and methylpiperidine derivatives (1 f, g) showed
a significantly decreased inhibitory activity. Decreasing the
spacer length from propylene to ethylene group resulted in
a decreased activity (compare 1a and 1h), whereas a butylene
spacer was as well tolerated as a propylene spacer (1j and 1i).
The replacement of the 7-methoxy substituent by a benzyloxy
group also led to potent inhibition, specifically with the pyrroli-
dine derivative 1k. The corresponding azepane derivative 1l
showed a strong quenching effect in the TruHits assay; hence
the IC50 value could not be accurately determined. Our docking
studies (Figure 6) indicated that the benzyl group is inserted in
the surface groove formed by Glu142, Asp189 and Lys148,
where it could undergo additional interactions with the pro-
Figure 6. Predicted binding modes for 7-benzyloxy derivatives 1k and 4s.
A) Predicted binding mode of 1k. B) Predicted binding mode of 4s. The
docking poses are shown in cyan, Spindlin1 residues involved in the interac-
tions are depicted in green, the protein surface is shown in light grey, and
H-bond interactions as yellow dashed lines.
tein binding pocket. This might explain the good potency ob-
served for the 7-benzyloxy derivatives.
Regarding the thioquinazolinone series 4a–w, it again be-
comes evident that the nature of the lysine-mimetic group has
a major impact on the activities of the compounds. Basic ali-
phatic moieties connected by an alkylene spacer (preferably
propylene or butylene) to the main heterocyclic scaffold are
more favored than more bulky lysine-mimetic groups, like pen-
tamethylpiperidine moieties. Similar to the 4-aminoquinazo-
lines, amide groups are not well tolerated as lysine mimetics.
&
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
8
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
The effect of the substituents at the 6- and 7-positions of
the quinazoline ring was explored by acquiring different 6,7-
unsubstituted derivatives as well as dimethoxy, diethoxy, 7-
benzyloxy and 7-N,N-dimetlyaminopropoxy derivatives. More-
over, the different terminal lysine-mimetic groups were investi-
gated. Notably, compounds lacking the alkoxy groups at the 6-
and 7-positions of the heterocyclic ring (4g–m) generally
showed weak to no inhibitory activity against Spindlin1. The
effect of an introduction of 6,7-dimethoxy groups was rather
ambiguous: While compound 4o bearing a diethylaminobuty-
lene side chain as a lysine-mimetic group showed a significant
inhibitory activity (IC50 =21 mm), the analogous derivative with
a shorter spacer, 4n, exhibited a decreased activity. Thus far,
the 6,7-diethoxy thioquinazolinone derivatives 4a and 4c were
still among the most active compounds of the series showing
IC50 values of 22.7 and 25 mm, respectively.
Analogous to the 4-aminoquinazolines, a considerable im-
provement of the inhibitory activity could be achieved by the
introduction of a benzyloxy group at position 7, where the 7-
benzyloxy-6-methoxythioquinazolinone derivatives 4q, 4r, and
4s showed IC50 values in the low micromolar range. Our dock-
ing studies predicted a similar binding mode to the 7-benzy-
loxy-4-aminoquinazolinethione derivatives, where the benzyl
moiety is embedded in the cavity formed by Asp189, Glu142,
and Lys148 (Figure 6).
omatic cage of Spindlin1, as also shown by our docking stud-
ies.
The obtained results clearly revealed that 4-aminoquinazoli-
nethiones and thioquinazolinones bearing a basic aliphatic
side chain represent a promising lead scaffold for Spindlin1 in-
hibitors. Structural derivatization of these scaffolds was there-
fore performed, which enabled us to investigate the SAR of
these series. It became evident that the choice of an appropri-
ate lysine-mimetic group is crucial for attaining an inhibitory
activity against Spindlin1. Generally, compounds bearing
a basic aliphatic moiety connected by a propylene or butylene
space to the quinazoline ring showed the best activities. An
improvement of the inhibitory activity could be achieved by
introducing a benzyloxy substituent at position 7 of the 4-ami-
noquinazolinethione or thioquinazolinone scaffold; several
benzyloxy derivatives inhibited Spindlin1 in low micromolar
concentrations, for example, derivatives 1k and 4q with IC50
values of 3.5 and 4.5 mm, respectively.
Up to now, the potent G9a inhibitor A366, which has been
discovered in our research group by focused screening of an
in-house database, still remains by far the most active inhibitor
reported for Spindlin1 (IC50 =186 nm). A366 therefore repre-
sents a highly promising scaffold for the development of
Spindlin1 inhibitors, and ongoing research is dedicated in our
laboratories to optimize its structure in order to achieve selec-
tivity over G9a. On the other hand, the herein described VS ap-
proach was successful in identifying several active hits, albeit
with moderate activities (IC50 of 23–86 mm). This comes as no
surprise, as it is well documented that the majority of success-
ful VS campaigns result in the identification of hits active in
the micromolar range.[35] Taking into consideration that Spind-
lin1 is a relatively new target with only one reported inhibitor
so far, the results obtained were satisfactory, especially as it en-
abled us to identify new scaffolds which can serve as lead
structures for novel Spindlin1 inhibitors, also in hybrids, for ex-
ample, with A366 fragments. Our attempts to structurally opti-
mize two of the identified lead scaffolds so far led to a four- to
six-fold improvement in the activities, with the most active de-
rivatives showing an IC50 values in the single-digit micromolar
range (~4 mm).
To address the electronegative surface groove of Spindlin1,
compounds containing an additional basic side chain at posi-
tion 7 were synthesized. Docking studies clearly indicated that
the additional basic amine moiety could undergo salt bridge
interactions with either Asp184 or Asp187 and Asp189.
However, this notion could not be substantiated by in vitro re-
sults; only compound 4u showed a moderate activity (IC50 =
34 mm) with no improvement over the ethoxy or benzyloxy de-
rivatives.
Conclusions
Spindlin1 is a H3K4me3 reader protein which was found to be
associated with several types of malignant tumors, including
ovarian cancer, liver carcinomas, non-small-cell lung cancers
and liposarcoma.[18] Small-molecule inhibitors which disrupt
the binding of the histone peptide to Spindlin1 should serve
as valuable chemical probes to investigate its biological roles
and assess its validity as a therapeutic target and may repre-
sent a potential novel therapeutic strategy in cancer treatment.
Here we applied a combination of computational techniques,
synthetic chemistry and in vitro assays in order to identify
small-molecule inhibitors of Spindlin1. With the help of a struc-
ture-based pharmacophore screening approach coupled with
docking studies, we were able to identify four compounds
which showed significant in vitro inhibition of Spindlin1 in the
applied AlphaLISA assay.
Experimental Section
Combined pharmacophore and docking-based screening
Commercial libraries: The ChemBridge and ChemDiv libraries (re-
lease 2013) were downloaded from the ZINC website[27] (zinc.dock-
ing.org) in SD format. The structures of the substances were
cleaned using MOE[28] (version 2012.10, Chemical Computing
Group) after conversion into MDB file format. The cleaning proce-
dure included salt stripping, protonation at pH 7 and energy mini-
mization. Subsequently, the molecular weight, number of H-bond
donors and number of positively charged N atoms was calculated
in MOE version 2012.10. Filtering was accomplished by removing
all compounds with a molecular weight >500 Da, with no H-bond
donors or no positively charged N atoms. A multi-conformational
dataset of the filtered libraries was generated using the idbgen
module implemented in LigandScout 3.1: The maximum number
of conformations was set to 25 (t=omega-fast).
A subsequent chemical similarity search to scan for close
structural analogues of the identified hits resulted in uncover-
ing four additional active hits. All the retrieved active inhibitors
of Spindlin1 possess an aliphatic basic group, which should act
as a lysine-mimetic and disrupt the binding of Kme3 to the ar-
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
9
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
Pharmacophore screening: The high resolution (2.1 ꢂ) crystal
structure of Spindlin1 with H3K4me3 residues 1–8 (PDB ID: 4H75)
was taken from the PDB. A preliminary pharmacophore model was
generated using the automatic pharmacophore generation proto-
col in LigandScout 3.1 to denote the pharmacophoric features for
the peptide-protein binding interactions. Residues of the protein
binding pocket were assigned as excluded volume features. The
model was manually curated: The hydrophobic feature generated
for the lysine aliphatic chain was removed as well as the H-bond
60 nm, respectively and pre-incubated for 10 min at room tempera-
ture. Stock solutions of the tested compounds (10 mm) were pre-
pared in DMSO. For IC50 determination experiments, serial 1:1 dilu-
tions of compounds in assay buffer were performed (final concen-
trations ranging from 0.1 to 200 mm) while maintaining the DMSO
concentration in all dilutions at 4%. Ten microliters 2ꢃ ligand solu-
tion were dispensed onto a 384 well AlphaPlateTM (PerkinElmer) in
triplicates. Subsequently, ten microliters of the His–Spindlin1(49–
262)/biotin–H3(1–23K4me3 solution were added to each well con-
taining ligand solution (AI) or 10 mL of 4% DMSO in assay buffer
(Apos, positive control). For the negative control (Aneg), 20 mL of
2% DMSO in assay buffer were added to the microplate. The final
concentrations were 15 nm His–Spindlin1(49–262) and 30 nm
biotin–H3(1–23K4me3 in a volume of 20 mL. The sealed microplate
was then incubated at 258C for 60 min. Streptavidin coated Al-
phaScreen donor beads and nickel chelate functionalized AlphaLI-
SA acceptor beads were diluted and mixed in assay buffer to
obtain 0.08 mgmLꢁ1 of each bead species. Five microliters of
mixed beads were added to each well on the microplate leading
to a final bead concentration of 0.017 mgmLꢁ1 per well. A total of
25 mL of 1.6% DMSO in assay buffer were dispensed into 16 wells
as blank controls. The sealed plate was further incubated at 258C
for 60 min in the dark due to the light sensitivity of the donor
beads. The readout was recorded using an EnVisionTM plate reader
(PerkinElmer) in a mode. Averages of all controls were calculated
after subtraction of blank values from the raw data. Inhibition
values (I) were calculated using the following Equation (1):
+
donor features assigned for Ala1 backbone NH3 and Arg2 side
chain, and only the assigned cationic feature (CAT-2 and CAT-3)
were kept for these residues. Based on the generated model, six
different 3-featured pharmacophore models were constructed by
alternatively enabling and disabling some features. The six con-
structed pharmacophore models were screened against the previ-
ously generated multiconformational databases using the iscreen
module implemented in LigandScout 3.1, where default settings
were used. The retrieved pharmacophore hits were saved in SD
format to be used for further docking studies.
Docking: The retrieved pharmacophore hits were energy minimized
in MOE (version 2012.10; Chemical Computing Group, Montreal,
Canada) using the MMFF94X force field. The crystal structure of
Spindlin1 (PDB ID: 4H75) was used for the docking study. The protein
structure was prepared with Schrçdinger’s Protein Preparation
Wizard:[36] Hydrogen atoms were added and the H-bond network
was subsequently optimized. The protonation states at pH 7.0 were
predicted using the PROPKA tool within the Schrçdinger program.
Only one water molecule (HOH416), located between Met140 and
Asp184, was kept and considered in the docking procedure. The
structure was finally subjected to a restrained energy minimization
step using the OPLS2005 force field (RMSD of the atom displacement
for terminating the minimization 0.3 ꢂ). Receptor grid preparation for
the docking procedure was carried out by assigning the K4me3 and
Thr3 residues of H3K4me3 as the centroid of the grid box. Molecular
docking was carried out using Glide[24] (Schrçdinger Inc., New York,
USA) in the Standard Precision mode. A total of 20 poses per
ligand were included in the post-docking minimization step and
a maximum of two docking poses were output for each ligand.
ꢀ
ꢀ
ꢁꢁ
AI ꢁ Aneg
I ¼ 100 ꢂ 1 ꢁ
½%ꢃ
ð1Þ
Apos ꢁ Aneg
GraphPad Prism was used for visualization and calculation of IC50
values by plotting inhibition values against the logarithmic com-
pound concentrations and applying a sigmoidal dose–response fit
(variable slope).
AlphaLISA TruHits verification assay
2D similarity search: The 2D similarity searches were performed
with MAACS key fingerprints using MOE.[28] The active compounds
1a, 2a and 3a were chosen as query structures. The prepared not
filtered ChemBridge and ChemDiv databases (previously described)
were screened, and a Tanimoto coefficient of 0.75 (query: 1a) or
0.80 (query: 2a, 3a) was set as a minimum value to fulfill the struc-
tural similarity requirement. The attained hits were subsequently
docked using the previously described settings.
The TruHits verification assay kit (PerkinElmer, cat. #AL900D) con-
tains AlphaLISAꢄ BSA–biotin acceptor beads and streptavidin func-
tionalized AlphaScreen donor beads which interact together to
generate an a signal without adding any further interaction part-
ners. A bead premix was prepared containing both bead species
diluted in assay buffer (0.019 mgmLꢁ1 of each bead type leading
to a concentration of 0.011 mgmLꢁ1 after addition to sample dilu-
tions in a final assay volume of 25 mL) which was pre-incubated at
room temperature for 30 min. A total of 10 mL of test compounds
was dispensed onto a 384-well Alpha plate in duplicates and the
pre-incubated bead premix (15 mL) was subsequently added. Thus,
the test compounds are brought to similar concentrations as in
the AlphaLISA assay. The microplate was sealed and incubated at
258C for 10 min. After spinning down the plate for 2 min at 300ꢃ
g, a signals were measured using the EnVisionTM plate reader as
described above. Potential quenching effects of the a signal were
determined by comparing the TruHits signal of positive controls
(Apos, only TruHits bead premix and buffer) to the signal of the
compound-bead mixtures (AI) after subtracting the blank values
(only buffer and no bead mixture). The amount of a signal quench-
ing (Q) was calculated by the following Equation (2):
Evaluation of drug-likeness, lead-likeness, and pan-assay interfer-
ence (PAINS) behavior: The drug- and lead-likeness of the VS hits
were calculated using MOE, where all the compounds showed <2 vi-
olations of the Lipinski’s rule-of-five[37] and the Oprea criteria of lead-
likeness[38] (lip_violation and opr_violations descriptors). Moreover, we
assessed whether the compounds possessed any problematic sub-
structural features which should render them unspecific and pan-
assay-interferers by applying the PAINS[29] filter. Screening against
PAINS was performed using the program KNIME,[39] using a workflow
downloaded
from
http://www.myexperiment.org/workflows/
1841.html.[40] None of the VS hits was recognized as potential PAINS.
AlphaLISA displacement assay
ꢀ
ꢁ
AI
Apos
Recombinant His–Spindlin1(49–262) and biotin–H3(1–23)K4me3
were mixed in assay buffer at fixed concentrations of 30 and
Q ¼ 100 ꢂ 1 ꢁ
½%ꢃ
ð2Þ
&
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
10
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
3.0 equiv) was added and a mixture of glacial acetic acid and
acetic anhydride was used as solvent. The reaction mixture was
stirred for 1 h at 08C and subsequently for 4 h at RT and finally
poured onto ice-water. After adjusting the pH to 8–10 using
Na2CO3, the obtained suspension was extracted with ethyl acetate
and washed with brine. The organic solvent was evaporated under
vacuum and the crude product was purified by gradient column
chromatography (ethyl acetate/cyclohexane, 1:9).
Synthesis
Starting materials and reagents were purchased in the highest
available purity grade and used as purchased. All solvents were
freshly distilled before use or purchased as HPLC-purity grade.
Thin-layer chromatography was carried out on aluminum sheets
coated with silica gel 60 F245 (Merck, Darmstadt, Germany). Flash-
column chromatography was carried out with the flash purification
system Isolera One (Biotage) using different silica gel cartridges (Bi-
otage SNAP KPSil cartridges (10 g, 25 g or 50 g) or Telos Silica
Flash-LL (12 g, 20 g, 40 g). Loaded cartridges were either directly
purchased or refilled with 40–63 mm silica gel 60 (Merck). A con-
centration gradient with two solvents was used to afford better
separation. Microwave assisted reactions were carried out in a CEM
Discover Microwave Synthesizer using a septum-closed reaction
vial (up to 5 mL) or closed reaction vessels (up to 20 mL). Tempera-
ture and pressure were continuously monitored.
Method C: General procedure for the reduction of the and
methyl nitrobenzoate (7) nitrobenzonitrile derivatives (12a,c): A
solution of the respective nitrobenzonitrile derivative (1.0 equiv) in
a mixture of ethanol/glacial acetic acid/water (2:2:1) was treated
with iron powder (5 equiv) and then in an ultrasound bath at 308C
until the complete conversion of the substrate as monitored by
TLC. The mixture was filtered off, poured into water and extracted
three times with ethyl acetate. The collected organic phase was
washed with brine and 0.5% citric acid solution, dried over Na2SO4
and evaporated under reduced pressure. The crude product was
purified by gradient column chromatography (ethyl acetate/cyclo-
hexane, 1:9).
Melting points were taken using Melt-Temp II (Laboratory Devices,
USA) and are uncorrected. NMR spectra were recorded on Bruker
Avance III HD spectrometer, 400 MHz for H- and 100 MHz for 13C-
1
spectra, using deuterated chloroform (CDCl3) or DMSO ((CD3)2SO).
Chemical shifts are reported in parts per million (d relative to the
residual solvent peak). The multiplicity of the signals is indicated
with the following abbreviations: bs (broad singlet), s (singlet), d
(doublet), t (triplet), q (quartet), m (multiplet), dd (doublet of dou-
blets), and td (doublet of triplets). Mass spectrometric analysis was
performed with Expression CMS (Advion) mass spectrometer for
the ESI-MS spectra and with an Exactive Spectrophotometer
(Thermo Scientific) for high-resolution ESI- and APCI-spectra. Com-
pound purity was assessed by LC using Agilent technology 1260
Infinity HPLC system with a UV detector at l=210 nm. It was oper-
ated at a flow rate of 1 mLminꢁ1 and the eluting solvents were
water (component A) and acetonitrile each containing 0.05% tri-
fluoroacetic acid (component B). A linear gradient elution was
chosen as following, min 1–4: Three methods were used with 90%
component A, 10% component B; min 4–35 90: to 30% compo-
nent A, 10% to 70% component B; min 35–40: 90%component A,
10% component B. Two different columns were used; Method A:
SynergiTM Polar-RP, 4 mm, 80 ꢂ 250ꢃ4.6 mm (Phenomenex) and
Method B: SynergiTM Hydro-RP, 4 mm, 80 ꢂ 250ꢃ4.6 mm (Phenom-
enex). The HPLC purity of all final compounds was ꢄ95%. Charac-
terization data of the synthesized compounds is given in the Sup-
porting Information.
Method D: General procedure for the synthesis of the alkyl chlo-
ride derivatives (12b) to the desired tertiary amines (12c) (Fin-
kelstein reaction): A mixture of the alkyl chloride derivative
(1 equiv), NaI (2.0 equiv) and tetrabutylammonium chloride (TBAI,
0.04 equiv) in acetone was irradiated at 40 W at 1008C for 10 min
in a microwave. The respective secondary amine (0.9 equiv) and
anhydrous K2CO3 were added, and irradiation in the microwave
was continued for 30 min at 1208C at 60 W. The mixture was
evaporated under reduced pressure, and the product was suspend-
ed in water and extracted three times with dichloromethane. The
collected organic phases were washed with a saturated NaHCO3
solution and then with brine, dried over Na2SO4 and evaporated
under reduced pressure. The crude product was purified by gradi-
ent column chromatography (dichloromethane/methanol/TEA,
99:0.95: 0.05).
Method E: General procedure for the conversion of the primary
aromatic amines (8a,b and 13a–c) to the corresponding isothio-
cyanates (9a,b and 14a–c): A mixture of thiophosgene (1.0 equiv)
and K2CO3 (2.0 equiv) in dichloromethane/water (3:7) was cooled
at 08C. An ice-cold solution of the corresponding primary aromatic
amine (1.2 equiv) in dichloromethane was then added dropwise to
the reaction mixture under rigorous stirring. The reaction mixture
was stirred at RT over-night, and finally the aqueous phase was ex-
tracted twice with dichloromethane. The organic phase was subse-
quently washed with brine, dried over Na2SO4 and evaporated
under reduced pressure. The crude product was purified by gradi-
ent column chromatography (ethyl acetate/cyclohexane, 5:95).
Method A: General procedure for the synthesis of the arylalkoxy
ether derivatives (6 and 11a,b): The respective bromo- or iodoal-
kane (1.5 equiv per one hydroxy group) was added dropwise to
a
stirred solution of the respective phenol (1 equiv), K2CO3
(3 equiv) and a catalytic amount of tetrabutylammonium bromide
(TBAB, 0.2 equiv) in acetone. The reaction mixture was then heated
under reflux for 2 h. After cooling down, the reaction mixture was
filtered and the filtrate was evaporated under vacuum. The crude
product was subsequently dissolved in ethyl acetate and washed
with brine and water. The organic phase was finally dried over
Na2SO4 and evaporated under vacuum. Final purification was ach-
ieved by gradient column chromatography (ethyl acetate/cyclohex-
ane, 1:9).
Method F: General procedure for the conversion of the methyl
2-isothiocyanato-benzoate (9a,b) and 2-isothiocyanato-benzoni-
trile derivatives (14a–c) into the quinazoline derivatives: A solu-
tion of the respective the 2-isothiocyanato-benzoic acid or 2-iso-
thiocyanato-benzonitrile derivatives (1.1 equiv), the respective
amine (1.0 equiv) and triethylamine (3.0 equiv) in dry THF was irra-
diated in a microwave at 1508C at 100 W for 45 min. The mixture
was cooled in the fridge overnight, and the resulting precipitate
was recrystallized from ethanol. In the cases where the product
failed to precipitate, the solvent was evaporated under reduced
pressure, and the crude product was purified by gradient column
chromatography (dichloromethane/methanol/TEA, 99:0.95: 0.05).
Method B: General procedure for the nitration of the benzoni-
trile (6) and methyl benzoate derivatives (11a,b): A solution of
the respective benzoic acid or benzonitrile derivatives (1 equiv) in
glacial acetic acid was cooled to 08C, and HNO3 (65%, 3.0 equiv)
was added dropwise under stirring. For the benzonitrile derivatives
a
cold mixture of H2SO4 (96%, 2.0 equiv) and HNO3 (65%,
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
11
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
C. H. Arrowsmith, G. Mer, K. M. McBride, L. I. James, S. V. Frye, ACS Chem.
Biol. 2015, 10, 1072–1081.
Acknowledgements
[15] T. Wagner, H. Greschik, T. Burgahn, K. Schmidtkunz, A. K. Schott, J. Mc-
Millan, L. Baranauskiene, Y. Xiong, O. Fedorov, J. Jin, U. Oppermann, D.
Matulis, R. Schule, M. Jung, Nucleic Acids Res. 2016, 44, e88.
[16] a) N. Milosevich, F. Hof, Biochemistry 2016, 55, 1570–1583; b) T. Wagner,
D. Robaa, W. Sippl, M. Jung, ChemMedChem 2014, 9, 466–483.
[17] L. Morera, M. Lubbert, M. Jung, Clin. Epigenet. 2016, 8, 57.
[18] a) H. Franz, H. Greschik, D. Willmann, L. Ozretic, C. A. Jilg, E. Wardel-
mann, M. Jung, R. Buettner, R. Schule, Oncotarget 2015, 6, 4773–4789;
b) F. Jiang, Q. Zhao, L. Qin, H. Pang, X. Pei, Z. Rao, Protein Pept. Lett.
2006, 13, 203–205; c) J. X. Wang, Q. Zeng, L. Chen, J. C. Du, X. L. Yan,
H. F. Yuan, C. Zhai, J. N. Zhou, Y. L. Jia, W. Yue, X. T. Pei, Mol. Cancer Res.
2012, 10, 326–335.
M.J. (Project A04) and R.S. (Project B1) thank the Deutsche For-
schungsgemeinschaft (DFG) [CRC992, Medical Epigenetics]. R.S.
further thanks the European Research Council [ERC AdGrant
322844], the DFG [CRC850, CRC746, and Schu688/12-1]. We
thank Richard T. Cummings and Nico Cantone (Constellation
Pharmaceuticals, Cambridge, MA, USA) for carrying out the
DOTL1 and EZH2 assays, and Holger Greschik for his guidance
with the Spindlin1 assays.
[19] a) H. Yuan, P. Zhang, L. Qin, L. Chen, S. Shi, Y. Lu, F. Yan, C. Bai, X. Nan,
D. Liu, Y. Li, W. Yue, X. Pei, Gene 2008, 410, 67–74; b) P. Zhang, B. Cong,
H. Yuan, L. Chen, Y. Lv, C. Bai, X. Nan, S. Shi, W. Yue, X. Pei, J. Cell. Physi-
ol. 2008, 217, 400–408.
Keywords: AlphaLISA assay · lead optimization · methyllysine
readers · Spindlin1 · virtual screening
[1] T. Kouzarides, Cell 2007, 128, 693–705.
[20] R. F. Sweis, M. Pliushchev, P. J. Brown, J. Guo, F. Li, D. Maag, A. M. Petros,
N. B. Soni, C. Tse, M. Vedadi, M. R. Michaelides, G. G. Chiang, W. N. Pap-
pano, ACS Med. Chem. Lett. 2014, 5, 205–209.
[21] N. Yang, W. Wang, Y. Wang, M. Wang, Q. Zhao, Z. Rao, B. Zhu, R. M. Xu,
Proc. Natl. Acad. Sci. USA 2012, 109, 17954–17959.
[22] J. M. Herold, L. A. Ingerman, C. Gao, S. V. Frye, Curr. Chem. Genomics
2011, 5, 51–61.
[23] X. Su, G. Zhu, X. Ding, S. Y. Lee, Y. Dou, B. Zhu, W. Wu, H. Li, Genes Dev.
2014, 28, 622–636.
[2] a) G. Egger, G. Liang, A. Aparicio, P. A. Jones, Nature 2004, 429, 457–
463; b) D. E. Handy, R. Castro, J. Loscalzo, Circulation 2011, 123, 2145–
2156.
[3] a) M. A. Dawson, T. Kouzarides, Cell 2012, 150, 12–27; b) S. Sharma, T. K.
Kelly, P. A. Jones, Carcinogenesis 2010, 31, 27–36.
[4] a) R. Lardenoije, A. Iatrou, G. Kenis, K. Kompotis, H. W. Steinbusch, D.
Mastroeni, P. Coleman, C. A. Lemere, P. R. Hof, D. L. van den Hove, B. P.
Rutten, Prog. Neurobiol. 2015, 131, 21–64; b) I. A. Qureshi, M. F. Mehler,
Mol. Aspects Med. 2013, 34, 875–882.
[5] a) J. M. Lorenzen, F. Martino, T. Thum, Basic Res. Cardiol. 2012, 107, 245;
b) T. Nꢀhrenberg, R. Gilsbach, S. Preissl, T. Schnick, L. Hein, Cell Tissue
Res. 2014, 356, 585–600.
[6] a) C. A. Musselman, M. E. Lalonde, J. Cote, T. G. Kutateladze, Nat. Struct.
Mol. Biol. 2012, 19, 1218–1227; b) S. D. Taverna, H. Li, A. J. Ruthenburg,
C. D. Allis, D. J. Patel, Nat. Struct. Mol. Biol. 2007, 14, 1025–1040.
[7] a) K. L. Yap, M. M. Zhou, Crit. Rev. Biochem. Mol. Biol. 2010, 45, 488–505;
b) M. Yun, J. Wu, J. L. Workman, B. Li, Cell Res. 2011, 21, 564–578.
[8] D. J. Patel, Z. Wang, Annu. Rev. Biochem. 2013, 82, 81–118.
[9] M. Brand, A. M. Measures, B. G. Wilson, W. A. Cortopassi, R. Alexander,
M. Hoss, D. S. Hewings, T. P. Rooney, R. S. Paton, S. J. Conway, ACS Chem.
Biol. 2015, 10, 22–39.
[10] a) M. A. Camerino, N. Zhong, A. Dong, B. M. Dickson, L. I. James, B. M.
Baughman, J. L. Norris, D. B. Kireev, W. P. Janzen, C. H. Arrowsmith, S. V.
Frye, MedChemComm 2013, 4, 1501–1507; b) J. M. Herold, T. J. Wigle,
J. L. Norris, R. Lam, V. K. Korboukh, C. Gao, L. A. Ingerman, D. B. Kireev,
G. Senisterra, M. Vedadi, A. Tripathy, P. J. Brown, C. H. Arrowsmith, J. Jin,
W. P. Janzen, S. V. Frye, J. Med. Chem. 2011, 54, 2504–2511; c) L. I.
James, D. Barsyte-Lovejoy, N. Zhong, L. Krichevsky, V. K. Korboukh, J. M.
Herold, C. J. MacNevin, J. L. Norris, C. A. Sagum, W. Tempel, E. Marcon,
H. Guo, C. Gao, X. P. Huang, S. Duan, A. Emili, J. F. Greenblatt, D. B.
Kireev, J. Jin, W. P. Janzen, P. J. Brown, M. T. Bedford, C. H. Arrowsmith,
S. V. Frye, Nat. Chem. Biol. 2013, 9, 184–191; d) L. I. James, V. K. Kor-
boukh, L. Krichevsky, B. M. Baughman, J. M. Herold, J. L. Norris, J. Jin,
D. B. Kireev, W. P. Janzen, C. H. Arrowsmith, S. V. Frye, J. Med. Chem.
2013, 56, 7358–7371; e) D. Kireev, T. J. Wigle, J. Norris-Drouin, J. M.
Herold, W. P. Janzen, S. V. Frye, J. Med. Chem. 2010, 53, 7625–7631.
[11] E. K. Wagner, N. Nath, R. Flemming, J. B. Feltenberger, J. M. Denu, Bio-
chemistry 2012, 51, 8293–8306.
[24] Schrçdinger Suite 2012: Glide Version 5.8, Schrçdinger LLC, New York,
NY (USA), 2012.
[25] H. M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H. Weissig,
I. N. Shindyalov, P. E. Bourne, Nucleic Acids Res. 2000, 28, 235–242.
[26] G. Wolber, T. Langer, J. Chem. Inf. Model. 2005, 45, 160–169.
[27] J. J. Irwin, T. Sterling, M. M. Mysinger, E. S. Bolstad, R. G. Coleman, J.
Chem. Inf. Model. 2012, 52, 1757–1768.
[28] Molecular Operating Environment (MOE) 2012.10, Chemical Computing
Group Inc., 1010 Sherbooke Street West, Suite 910, Montreal, QC,
H3A 2R7 (Canada), 2012.
[29] J. B. Baell, G. A. Holloway, J. Med. Chem. 2010, 53, 2719–2740.
[30] A. Williamson, Ann. Chem. Pharm. 1851, 77, 37–49.
[31] A. B. Gamble, J. Garner, C. P. Gordon, S. M. J. O’Conner, P. A. Keller, Synth.
Commun. 2007, 37, 2777–2786.
[32] G. Calestani, L. Capella, R. Leardini, M. Minozzi, D. Nanni, R. Papa, G. Za-
nardi, Tetrahedron 2001, 57, 7221–7233.
[33] a) S. E.-S. Barakat, M. M. Ghorab, H. M. Saker, M. M. Abd Rabo, Phospho-
ˇ
rus Sulfur Silicon Relat. Elem. 2007, 182, 65–77; b) J. Rousseau, V. Krisciu-
ˇ
niene, I. Rimkeviciute, C. C. Rousseau, V. Amankaviciene, A. Sackus, A.
Tatibouꢅt, P. Rollin, Monatsh. Chem. 2009, 140, 339–348.
[34] S. Garapaty-Rao, C. Nasveschuk, A. Gagnon, E. Y. Chan, P. Sandy, J.
Busby, S. Balasubramanian, R. Campbell, F. Zhao, L. Bergeron, J. E.
Audia, B. K. Albrecht, J. C. Harmange, R. Cummings, P. Trojer, Chem. Biol.
2013, 20, 1329–1339.
[35] P. Ripphausen, B. Nisius, L. Peltason, J. Bajorath, J. Med. Chem. 2010, 53,
8461–8467.
[36] Schrçdinger Suite 2012: Protein Preparation Wizard, Epik Version 2.3,
Impact Version 5.8, Prime Version 3.1; Schrçdinger LLC, New York, NY
(USA), 2012.
[37] C. A. Lipinski, F. Lombardo, B. W. Dominy, P. J. Feeney, Adv. Drug Delivery
Rev. 2001, 46, 3–26.
[38] T. I. Oprea, J. Comput.-Aided Mol. Des. 2000, 14, 251–264.
[39] M. Berthold, N. Cebron, F. Dill, T. Gabriel, T. Kotter, T. Meinl, P. Ohl, K.
Thiel, B. Wiswedel, ACM SIGKDD Explorations Newsletter 2009, 11, 26–
31.
[12] T. C. Miller, T. J. Rutherford, K. Birchall, J. Chugh, M. Fiedler, M. Bienz,
ACS Chem. Biol. 2014, 9, 2864–2874.
[13] a) C. Ren, K. Morohashi, A. N. Plotnikov, J. Jakoncic, S. G. Smith, J. Li, L.
Zeng, Y. Rodriguez, V. Stojanoff, M. Walsh, M. M. Zhou, Chem. Biol. 2015,
22, 161–168; b) C. Ren, S. G. Smith, K. Yap, S. Li, J. Li, M. Mezei, Y. Rodri-
guez, A. Vincek, F. Aguilo, M. J. Walsh, M. M. Zhou, ACS Med. Chem. Lett.
2016, 7, 601–605; c) J. I. Stuckey, B. M. Dickson, N. Cheng, Y. Liu, J. L.
Norris, S. H. Cholensky, W. Tempel, S. Qin, K. G. Huber, C. Sagum, K.
Black, F. Li, X. P. Huang, B. L. Roth, B. M. Baughman, G. Senisterra, S. G.
Pattenden, M. Vedadi, P. J. Brown, M. T. Bedford, J. Min, C. H. Arrow-
smith, L. I. James, S. V. Frye, Nat. Chem. Biol. 2016, 12, 180–187.
[14] M. T. Perfetti, B. M. Baughman, B. M. Dickson, Y. Mu, G. Cui, P. Mader, A.
Dong, J. L. Norris, S. B. Rothbart, B. D. Strahl, P. J. Brown, W. P. Janzen,
[40] S. Saubern, R. Guha, J. B. Baell, Mol. Inf. 2011, 30, 847–850.
Received: July 18, 2016
Revised: August 22, 2016
Published online on && &&, 0000
&
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
12
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

FULL PAPERS
D. Robaa, T. Wagner, C. Luise, L. Carlino,
J. McMillan, R. Flaig, R. Schꢀle, M. Jung,
W. Sippl*
&& – &&
Identification and Structure–Activity
Relationship Studies of Small-Molecule
Inhibitors of the Methyllysine Reader
Protein Spindlin1
Stopping the reader: An in silico
recognized as promising lead structures.
Subsequent chemical optimization was
carried out, and a series of derivatives
of both lead scaffolds were synthesized.
This resulted in the discovery of Spind-
lin1 inhibitors and helped explore the
SAR of these inhibitor series.
screening approach coupled with in
vitro testing was performed to search
for inhibitors of the methyllysine reader
protein Spindlin1. Among the eight
identified inhibitors, 4-aminoquinazoline
and quinazolinethione derivatives were
ChemMedChem 2016, 11, 1 – 13
www.chemmedchem.org
13
ꢁ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ