Luciferin and derivatives as a DYRK selective scaffold for the design of protein kinase inhibitors

Article abstract of DOI:10.1016/j.ejmech.2015.02.035

D-Luciferin is widely used as a substrate in luciferase catalysed bioluminescence assays for in vitro studies. However, little is known about cross reactivity and potential interference of D-luciferin with other enzymes. We serendipitously found that firefly luciferin inhibited the CDK2/Cyclin A protein kinase. Inhibition profiling of D-luciferin over a 103-protein kinase panel showed significant inhibition of a small set of protein kinases, in particular the DYRK-family, but also other members of the CMGC-group, including ERK8 and CK2. Inhibition profiling on a 16-member focused library derived from D-luciferin confirms that D-luciferin represents a DYRK-selective chemotype of fragment-like molecular weight. Thus, observation of its inhibitory activity and the initial SAR information reported here promise to be useful for further design of protein kinase inhibitors with related scaffolds.

Full text of DOI:10.1016/j.ejmech.2015.02.035

Accepted Manuscript
Luciferin and derivatives as a DYRK selective scaffold for the design of protein kinase
inhibitors
Ulli Rothweiler, Jonas Eriksson, Wenche Stensen, Frederick Leeson, Richard A.
Engh, John S. Svendsen
PII:
S0223-5234(15)00131-2
10.1016/j.ejmech.2015.02.035
EJMECH 7717
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 19 December 2014
Revised Date: 19 January 2015
Accepted Date: 19 February 2015
Please cite this article as: U. Rothweiler, J. Eriksson, W. Stensen, F. Leeson, R.A. Engh, J.S. Svendsen,
Luciferin and derivatives as a DYRK selective scaffold for the design of protein kinase inhibitors,
European Journal of Medicinal Chemistry (2015), doi: 10.1016/j.ejmech.2015.02.035.
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
D-Luciferin is an inhibitor of several protein kinases
The DYRK-family is particularly affected by a lucifein-library
X-ray structures of CDK2/inhibitor complexes reveal important binding interactions
Luciferin is an inhibitor scaffold with intrinsic selectivity for DYRK-kinases

ACCEPTED MANUSCRIPT
Luciferin and derivatives as a DYRK selective
scaffold for the design of protein kinase inhibitors
Ulli Rothweiler1, Jonas Eriksson2,†, Wenche Stensen2,3, Frederick Leeson2,3, Richard A.
Engh1,* and John S. Svendsen2,3,*
1The Norwegian Structural Biology Centre, Department of Chemistry, UiT The Arctic
University of Norway, N-9037 Tromsø, Norway, 2Lytix Biopharma AS, P.O. Box 6447,
Tromsø Science Park, N-9294 Tromsø, Norway and 3Department of Chemistry, UiT The
Arctic University of Norway, N-9037 Tromsø, Norway.
AUTHOR INFORMATION
Corresponding Authors
*Corresponding authors: richard.engh@uit.no +47 776 44073; john-sigurd.svendsen@uit.no
+47 776 44086.
Present Addresses
†Laboratories for Chemical Biology Umeå (LCBU), Chemical Biology Consortium Sweden
(CBCS), Department of Chemistry, Umeå University, SE-901 87 Umeå.
Keywords: luciferin, protein kinase, inhibitor profiling, drug design, crystallography
1

ACCEPTED MANUSCRIPT
Abstract
D-Luciferin is widely used as a substrate in luciferase catalyzed bioluminescence assays for
in vitro studies. However, little is known about cross reactivity and potential interference of
D-luciferin with other enzymes. We serendipitously found that that firefly inhibited the
CDK2/Cyclin A protein kinase. Inhibition profiling of D-luciferin over a 103-protein kinase
panel showed significant inhibition on a small set of protein kinases, in particular the DYRK-
family, but also other members of the CMGC-group, including ERK8 and CK2. Inhibition
profiling on a 16-member focused library derived from D-luciferin confirms that D-luciferin
represents a DYRK-selective chemotype of fragment-like molecular weight. Thus
observation of its inhibitory activity the initial SAR information reported here promise to be
useful for further design of protein kinase inhibitors with related scaffolds.
2

ACCEPTED MANUSCRIPT
1. Introduction
Novel compounds with protein kinase inhibitory properties may be highly valuable, and
screening for such activity is therefore included in many bioprospecting efforts. The kinase
RR[1] technology ("Reaction Rate") is a popular choice to identify fractions with protein
kinase inhibitory activity; it was developed to enable real-time monitoring target
peptide/protein phosphorylation by protein kinases.[2] In this assay, luciferase and D-luciferin
are present during the entire kinase reaction, allowing continuous measurement of light
emission, although for high throughput screening, measurement of light emission is required
only at the beginning and after a pre-set time point. In the late evaluation process prior to the
launch of Kinase RR, the assay was tested against a panel of 38 protein kinases selected to
represent active site diversity. Although Kinase RR performed well and could be consistent
against a reference assay method, there was one protein kinase that repeatedly failed; caseine
kinase 2 (CK2). During the Kinase RR implementation in a bioprospecting screen, a similar
failure was found for CDK2/Cyclin A. The reason for this failure could be attributed to be the
direct inhibitory action of D-luciferin on the CDK2/Cyclin A complex. The finding that many
protein kinases are unaffected by luciferin while a few kinases are inhibited by the compound
prompted us to investigate whether D-luciferin inhibited additional protein kinases, and
whether D-luciferin could serve as a scaffold for the construction of novel selective protein
kinase inhibitors, by preparing a luciferin derived focused library and profiling this library
against a large protein kinase panel.
3

ACCEPTED MANUSCRIPT
2. Results
2.1 Preparation of inhibitors and protein kinase activity studies
Kinase profiling assays of D-luciferin (1) at a concentration of 100 µM were performed with
a 103-kinase panel at the International Centre for Kinase Profiling at the University of
Dundee, UK.[3] The profile revealed that 21 kinases were inhibited significantly (≤ 25%
remaining activity) by D-luciferin, but also that the majority of kinases were unaffected even
at this relatively high concentration. Among the most luciferin sensitive kinases were the
DYRK family (1a, 2 and 3 tested), Auroras B and A, CK2, PKBβ, VEGF-R GCK, RSK2,
ERK8 and PIM3. The discovery that several kinases of potential interest as drug targets were
inhibited by D-luciferin prompted us to prepare an exploratory structure-activity relationship
(SAR) study on D-luciferin as a scaffold for design of kinase inhibitors (Figure 1).
(Insert Figure 1.)
The SAR-library design was based on the assumption that the functional groups, i.e. the
phenolic hydroxyl group on the benzothiazole ring system and the carboxylic acid on the
dihydrothiazole ring of luciferin, were important for binding interactions with the affected
kinases. These groups were substituted by methyl ether and methyl ester derivatives,
respectively. Further variations included the replacement of the bicyclic benzothiazole by a
quinoline, position of the hydroxyl group on the quinoline ring, the introduction of methyl
groups in the dihydrothiazole ring, and the absolute configuration of the carboxylic acid
substituent of the dihydrothiazole ring. The benzothiazole and quinoline SAR-libraries were
prepared in the conventional manner as outlined in Scheme 1 (following White[4]) by
condensing a 2-cyanobenzothiazole (or a 2-cyanoquinoline) with (R)- or (S)-cysteine (or (R)-
or (S)-penicillamine), with subsequent esterification of the carboxylic acid where necessary.
This method is used in nearly all preparations of lucifeins today.[5] The advantage of
4

ACCEPTED MANUSCRIPT
condensing cysteine derivatives with cyanobenzothiazoles or cyanoquinolines at a late stage
in the process is that the process is mild, produces high yields and preserves the stereocenter
of the cysteine derivative in the dihydrothiazole ring.[5] The preparative chemistry of D-
luciferin has been recently reviewed.[6]
The phenols, 6-hydroxy-1,3-benzothiazole-2-carbonitrile and 6-hydroxyquinoline-2-
carbonitrile, were prepared from their corresponding commercially available methyl ethers by
treatment with dry pyridinium chloride at 200 °C.[7] The individual members of this SAR-
library are presented in Table 1.
(Insert Scheme 1)
5

ACCEPTED MANUSCRIPT
Table 1. Structure and corresponding protein kinase inhibitory activity of compounds 1-16.
Inhibitory activity is expressed as % remaining kinase activity at an inhibitor concentration of
100 µM.
1 – 12
13 - 16
Compound
X
R
Y
*
DYRK2 DYRK1a DYRK3 CK2
Aurora
A
Aurora
B
CDK2/
Cyclin
A
OH
OH
OH
OH
H
H
H
H
H
S
4
9
4
14
8
5
7
16
64
96
42
76
81
43
81
57
37
58
26
75
95
60
78
6
75
91
84
82
106
83
85
109
99
77
88
7
39
43
29
33
92
91
33
89
59
46
29
12
64
66
45
45
1
2
H
R
S
10 41
30 45
32 52
55 84
55 83
15 66
18 65
21 12
CH3
CH3
H
13
14
31
34
15
14
10
3
3
R
S
9
4
OCH3
OCH3
OCH3
OCH3
H
H
H
H
24
20
41
32
27
5
5
H
R
S
6
CH3
CH3
H
7
R
S
8
OCH3 CH3
OCH3 CH3
9
H
R
S
4
9
5
5
10
41
21
9
10
11
12
13
14
15
16
OH
OH
CH3
CH3
H
4
4
H
R
S
1
2
6-OH H
6-OH H
8-OH H
8-OH H
H
4
8
H
R
S
17
5
21
37
52
25 34
14 14
49 52
21
5
H
H
R
27
6
dark grey: ≤ 15, light grey: ≤25
6

ACCEPTED MANUSCRIPT
2.2 Kinome Profiling Analysis
Finally, the remaining 15 members of the SAR-library were subjected to kinase inhibition
profiling using the same 103-kinase panel. All assays were performed with an ATP
concentration at or below Km for the particular protein kinase. The complete results from the
kinome profiling measured as per cent remaining protein kinase activity at a 100 M
inhibitor concentration are given in the Supplementary Information (SI) section, and the
inhibitory activity against a selected set of kinases is presented in Table 1. Strong inhibition
was defined as < 15% remaining activity, and significant inhibition was defined as being in
the range of 16% – 25% remaining activity. The Table in SI is sorted according to the
average of the remaining activities for a particular protein kinase over all 16 compounds in
the luciferin library. The protein kinases on the top of the Table are thus protein kinases that
are strongly inhibited by many members of the luciferin library. The protein kinase that is
most strongly inhibited across the luciferin library is DYRK2, with an average residual
activity of less than 13%. The two other DYRK family kinases represented in the profiling
set, DYRK1a and DYRK3, follow next on the list. PIM3, ERK8 and CK2 are three non-
DYRK protein kinases that are strongly inhibited by many compounds in the luciferin library.
Following these six kinases, the inhibition either becomes weaker or more irregular. Aurora
B is also strongly inhibited by a few compounds in the library, but is virtually unaffected by
the rest. Furthermore, a few kinases of established drug targeting interest are strongly
inhibited by a single member of the library, such as PKBβ by compound 10 and PKCζ by
compound 2. A subset of the full Table (SI), emphasizing the most strongly and broadly
inhibited kinases, is reproduced as Table 1 in the main text.
7

ACCEPTED MANUSCRIPT
2.3 Structure-Activity Relationships
Most of the SAR-trends were kinase specific, but masking of the phenol moiety as a methoxy
group generally reduced inhibitory activity compared with the corresponding phenolic
compound across the kinase set (with DYRK1a as an exception, being slightly less affected
than the other kinases). Esterification of the carboxylate generally reduced inhibitory activity
for the phenolic compounds (e.g. compounds 3 and 4 versus compounds 1 and 2), however
for the methoxy-compounds esterification increased activity (e.g. compounds 7 and 8 versus
compounds 5 and 6), in particular for DYRK2 and DYRK3. Introduction of two methyl
groups in the dihydrothiazole ring increased inhibitory activity, in particular for the methoxy-
compounds, rendering the methoxy-gem-dimethyl compounds 9 and 10 almost as effective as
their phenolic counterparts 11 and 12. The stereochemistry of the carboxylate moiety is
however more important for the methoxy-compounds 9 and 10, whereby the S-enantiomer 9
is significantly more active than the R-enantiomer 10. This effect is greatly attenuated for the
enantiomeric phenols 11 and 12. Otherwise, stereochemistry seems to play a surprisingly
small role for the protein kinases that are promiscuously inhibited by members of the
luciferin library. For the protein kinases that are more sensitive to scaffold modifications,
such as CK2 and Aurora B, the effect of carboxylate stereochemistry can be significant (e.g.
compounds 1 and 2) or irrelevant (compounds 9 and 10). There are only minor changes in the
kinase inhibition pattern when the 2-benzothiazole-system is exchanged with a 2-quinoline
system (e.g. compound 1 versus compound 13). The effect of carboxylate stereochemistry is
however more significant for the 6-hydroxyquinoline derivatives than for corresponding 6-
hydroxybenzothiazole derivatives. Moving the hydroxyl-group from the 6-position to the 8-
position in the quinoline ring system created more marked changes. The 8-hydroxy-
derivatives were generally more active, in particular in relation to DYRK 1a and DYRK3.
8

ACCEPTED MANUSCRIPT
2.4 Interaction with CDK2
The inhibitory action of D-luciferin on protein kinases was first discovered on CDK2/Cyclin
A. Our laboratory has a high throughput screening of kinase structure by soaking on
preformed CDK2 crystals, and the initial work and the structural studies were performed on
this system.
2.4.1 Thermal shift assay
A thermal shift assay was performed to see whether a binding of D-luciferin (1) to CDK2
could be detected by a change in melting temperature. The average thermal shift of the
melting temperature was increased by 5 °C for the CDK2 kinase in the presence of D-
luciferin. When the benzothiazole ring was replaced by a quinoline ring system as in
compounds 13 and 15, the thermal shift was significantly reduced (Table 2).
Table 2. Melting temperature (mT) in °C of CDK2 and complexes of CDK2 with inhibitors
1, 13 and 15.
Compound
mT
mT
0
mT(DMSO)
blank
DMSO
13
42.7
41.5
43.4
43.9
47.7
-
-1.2
+0.7
+1.2
+5
0
+1.9
+2.4
+6.2
15
1
2.4.2 X-ray crystal structures of CDK2 / inhibitor complexes
To investigate the binding of D-luciferin and its derivatives to CDK2, a set of crystallization
experiments were performed. Crystals of apo-CDK2 were soaked with different derivatives
and measured at the BESSY II Synchrotron in Berlin. The statistics of the structures are
summarized in Table 3. Complexation with compounds 1 and 15 were clearly visible in the
9

ACCEPTED MANUSCRIPT
electron density maps. Compound 15 had the best fit in the difference density, as seen in
Figure 2. Weaker difference density showed the position of compound 1, which could be
modelled into the map based on the quality of fit to the three strong difference density peaks,
attributed to the two sulphur atoms and the hinge binding OH-group. Compound 13 showed
unclear density for the fragment in the ATP pocket of CDK2. Compound 13 has only one
sulphur atom and the difference density peak for this sulphur atom is weak compared to
compound 1. Inhibitor 13 can still be modelled into the map based on the structure of the
CDK2/compound 1 complex, but the CDK2/compound 13 complex is not completely
covered by electron density even after refinement. The density map suggests rather that the
ATP pocket in this crystal has mixed occupancy, with two partial occupancy water molecules
binding to the hinge in addition to the partial occupancy inhibitor.
(Insert Figure 2.)
Compound 1 and 15 bind in different orientations in the ATP pocket of CDK2, as a result that
can be attributed to the difference in the OH location in the benzothiazole and quinoline ring
systems, respectively. For compound 15, the shift of the hydroxyl group to the 8-position of
the quinoline ring results in a rotation of the inhibitor scaffold by approximately 90 degrees
relative to D-luciferin (Figures 3 and 4), and a different residue in the hinge region is used for
hydrogen bonding to the hydroxyl group. Compound 15 creates a hydrogen bond to the
carbonyl of the main chain residue L83 (Gatekeeper+3) whereas compound 1 interacts
similarly to the carbonyl from residue E81 (Gatekeeper+1). The hydroxyl oxygen atom in
both inhibitors accept a hydrogen bond from the amide of L83.
(Insert Figure 3.)
10

ACCEPTED MANUSCRIPT
Compound 1 has a length suitable to form an additional hydrogen bond between the
carboxylate group and residue T14 in the glycine rich loop. Compound 15 penetrates deeper
into the pocket, due to the change in the overall orientation, and therefore does not bind to the
glycine rich loop, but makes a hydrogen bond to D145 instead. In Apo CDK2, D145 shares a
hydrogen bond with K33 and this H-bond is disrupted by inhibitor 15 binding (Figure 4).
(Insert Figure 4.)
Compound 1 does not displace this interaction, but instead makes hydrogen bonds with both
K33 and D145. A second difference between the binding modes of compound 1 and 15 is the
orientation of the dihydrothiazole ring. For compound 1, the dihydrothiazole sulphur atom is
oriented towards the solvent, whereas for compound 15 the dihydrothiazole ring is flipped
180 degrees and the sulphur atom is oriented towards the inside of the ATP pocket (Figure 3).
The presence of weak but significant difference density (Figure 2, Panel C) adjacent to the
sulphur atom in the dihydrothiazole ring of compound 1 might however indicate partial
occupancy of the carboxylate also at this position along with an orientation of the
dihydrothiazole ring similar to that of 15. Except for the differences in the binding pocket, the
overall structures of the kinase in complexes CDK2/1 and CDK2/15 are very similar. In the
CDK2/15 structure however, a second binding site for compound 15 in the C-lobe exists that
is not present in the CDK2/1 structure (Figure 5).
(Insert Figure 5.)
11

ACCEPTED MANUSCRIPT
Table 3. Data collection and refinement statistics
DATA COLLECTION
PDB code
Synchrotron radiation
Wavelength, Å
Number of crystals used
Number of frames
Oscillation range / frame
CDK2/1
4D1X
BESSY II
0.9184
1
CDK2/15
4D1Z
BESSY II
0.9184
1
130
1°
100
1°
DIFFRACTION DATA
Space group
P212121
P212121
Unit cell parameters, Å
Protein molecules in
asymmetric unit
53.31, 72.12, 72.29
53.83, 72.41, 73.06
1
1
Number of measurements
Unique reflections
Resolution Range, Å
(final shell)
94018
18036
95708
24735
50.0-2.1 (2.15-2.10)
50.0-1.85 (1.89-1.85)
Completeness (final shell)
I/σ (final shell)
99.7 (98.6)
19.62 / 3.72
99.0 (96.7)
15.9 / 3.4
Merging
(final shell)
R factor observed
6.7% / 46.0%
4.9% / 37.8%
REFINEMENT
Resolution limits, Å
(final shell)
42.9-2.1 (2.15-2.10)
30.2-1.85 (1.89-1.85)
Number of used reflections
Percentage observed
% of free reflections
Number of protein atoms
Number of heterogen atoms
Number of solvent atoms
R factor overall/free
Wilson B-factor
16820
99.9
5.1
2162
18
117
0.194/0.238
28.6 Å2
0.004Å / 0.86°
97% / 3%
22727
99.0
5.1
2016
38
145
0.195/0.225
25.2 Å2
0.005 Å / 0.97°
98.4% / 1.6%
RMS bonds/angles
Ramachandran
(favored / allowed)
3. Discussion
3.1 Structural interpretation of CDK2-SAR
CDK2/Cyclin A is only modestly inhibited by the compounds of the SAR-library. There are
however clear variations in the inhibitory activities for each member in the library due to the
change of structural features (Table 1), and most of these trends coincide, albeit at a more
attenuated level, with the trends observed for the more strongly inhibited protein kinases. The
crystal structures provide insight into key features of binding the luciferin derivatives to
CDK2 (Figures 3 and 4). The general loss of binding efficacy when the methoxy-group
12

ACCEPTED MANUSCRIPT
replaces the 6-hydroxy group is due to the loss of a hydrogen bond. The phenolic hydroxyl
group in D-luciferin (1) group anchors the molecule at the kinase hinge segment via two
hydrogen bonds: a hydrogen donor interaction to the carbonyl group of E81 (at the
Gatekeeper+1 site) with a close O-O distance of 2.8 Å, and a hydrogen acceptor interaction
with the amino group of L83 (Gatekeeper+3), with an O-N distance of 3.0 Å. An identical
binding orientation for an inhibitor with a methoxy group substitution (e.g. compound 5)
would not be possible due to steric repulsion from the methyl group. Furthermore, even a
displaced methoxy group could function only as a weaker hydrogen bond acceptor.[6] This
could explain the weaker binding of compounds 5, 6, and 8 compared to compounds 1-4.
However, the exceptional tighter binding of compound 7, despite the methoxy substitution,
requires further studies for clarification. Increase of the lipophilic steric bulk in the
dihydrothiazole ring system as in the penicillamine derivatives 9 – 12 leads to more effective
CDK2/Cyclin A inhibitors, a feature that may arise from solvophobic effects introduced by
the geminal dimethyl moiety.
On the other hand, the gain in inhibitory efficacy by moving from the carboxylates (e.g.
compound 1) to an ester moiety (e.g. compound 3) is not trivially explained. In the crystal
structure of complex CDK2/1, there is an important interaction between the carboxylate and
the salt bridge K33/D145, and it is expected that an esterification of the carboxylate would
diminish this interaction. However, the inhibition profile studies are performed not on CDK2
alone, but on the CDK2/Cyclin A complex. Complexation of CDK2 with Cyclin A changes
the conformation of the active site of CDK2 in accord with its function as activating
cofactor.[8, 9] Of particular importance is the movement of helix C (the PSTAIRE-helix) into
the catalytic cleft. This movement breaks the salt bridge between K33/D145 and introduces a
new salt bridge between K33 and E51.[10] The carboxylate moiety of D145 will be left
exposed to the carboxylate moiety of the luciferin derivatives and create a potential charge
13

ACCEPTED MANUSCRIPT
clash. This charge repulsion between D145 and the luciferin ligands will be reduced in the
carboxylic ester derivatives, and hence explain their higher efficacy as inhibitors.
Based on the understanding of the CDK2 structures and overall SAR data reported here, the
SAR features for other individual kinases may be analysed based on published structures.
3.2 SAR of DYRK-family inhibition
The members in the DYRK-family of the protein kinases are the most strongly inhibited
enzymes by the compounds in the luciferin-library. DYRK2 is the most sensitive kinase in
the DYRK-family and will serve as a model for the discussion. DYRK1a and DYRK3 are
less affected by the inhibitors, but the inhibition pattern is similar to that of DYRK2, with
minor differences. Protein kinases in the DYRK-family are members of the CMGC-kinases,
and kinases from other families in the CMGC-group are also significantly inhibited, albeit to
a lesser degree, or by fewer members of the library. ERK 8 (a MAP-kinase), CK2 and CDK2
are examples of this.
DYRK2 is promiscuously inhibited by most compounds in the luciferin SAR-library, with
only the methoxy carboxylates (compounds 5 and 6) as inactive inhibitors. Many DYRK-
family kinase inhibitors have been reported in the literature (see e.g. [11] for a recent review).
These falls into several diverse molecular frameworks including natural compounds such as
β-carbolines[3] and acridone[12] alkaloids, flavanols,[13] meridianines,[14] and
lamellarines.[15] Furthermore, synthetic natural product derivatives[16] and fully synthetic
molecular scaffolds[17-22] have been developed as kinase inhibitors. Despite this wealth of
inhibitor scaffolds, only a few X-ray crystallographic structures of inhibitor/DYRK kinase
complexes are available. These include complexes between DYRK1a and the benzothiazole
14

ACCEPTED MANUSCRIPT
INDY (3ANQ)[23], the β-carboline harmine[23] as well as two pyrido-pyrimidines (4MQ1
and 4MQ2),[24] and DYRK2 complexes with a bis-indole indigoid (3KVW)[25] or
Leucettine L41 (4AZF and 4AZE).[26] A superposition of published DYRK/inhibitor crystal
structures with the CDK2/1 complex (Figure 6) highlights several features, in particular that
the volume of the active site occupied by the inhibitors are distinctly different. As shown in
Figure 6, the inhibitors of the published DYRK1a complex structures interact via a phenolic
hydroxy- or a methoxy-group from the inhibitor with the hinge region L241 at the Gatekeeper
+3 position[23] in a manner similar to the interaction found in the CDK2/1 complex (vide
supra). Furthermore, the inhibitors interact through a carbonyl or a pyridine nitrogen with the
conserved ATP-binding K188 of DYRK1a in a manner similar to the interaction between the
carboxylate of the luciferin inhibitors and K33 of CDK2. Likewise, in the DYRK2 complex
with an indigoid inhibitor, similar interactions with the hinge backbone residues and the
conserved lysine are observed. Despite this similarity in interaction sites between the
published DYRK complexes and the CDK2/1 complex described here, Figure 6 clearly
shows that the placement and orientation of the DYRK inhibitors differs significantly from 1
in CDK2. This effect can be related to the different shape of the ATP-pocket in the DYRK
kinases and CDK2.
The differences in surfaces surrounding the pocket may be attributed to a combination of
sequence differences and possibly resultant conformation differences. At the C-lobe side of
the binding pocket (in this orientation, at the base), residues isoleucine (DYRK2) and leucine
(DYRK1a) restrict the volume relative to alanine of CDK2. Elsewhere at the base of the
binding pocket, mutually conserved residues from the C-lobe adopt different rotamers. On the
left, differing volumes and conformations of the Gatekeeper +2 residues (phenylalanine and
methionine in CDK2 and Dyrk1A, respectively) restrict the volume in CDK2. This is linked
to a structuring edge-face aromat-aromat interaction between the phenylalanine and the
15

ACCEPTED MANUSCRIPT
Gatekeeper +4 residue (histidine), which also induces a change in the hinge main chain
conformation. Other differences in the shape of the ATP pocket arise from different
conformations of the glycine rich loop, which may be due to inhibitor interactions and/or
sequence differences; DYRK kinases have a phenylalanine as aromatic residue at the beta
hairpin turn of the glycine-rich loop, while CDK2 has a tyrosine. Finally, the binding pockets
differ strongly due to positions of the catalytic lysine.
In CDK2, the kinase active site lysine K33 shares a salt bridge interaction with D145,
creating the pillar like surface at the rear of the pocket and blocking contact to the gatekeeper
phenylalanine. This may be an inhibitor-induced conformation, but seems most likely linked
to the absence of a cyclin in the CDK2 structure (vide supra). These differences in the shape
of the active site of DYRK2 and CDK2 do not offer an obvious explanation for why DYRK2
is more effectively inhibited by members of the luciferin SAR-library than CDK2/Cyclin A.
(Insert Figure 6.)
3.3 SAR of CK2 inhibition
CK2 has been a drug target for well over a decade, and many inhibitors have been identified,
one of which has gone to clinical trials.[27] The inhibitors are chemically highly diverse, but
many do possess a terminating carboxylic acid group reminiscent of the inhibitor series we
report here. A superposition of these inhibitors using coordinates deposited in the PDB show
that the CK2 carboxylate groups form one predominating cluster as salt bridge partners of the
active site lysine (Figure 7). D-luciferin (1) and the quinoline cognate 13 both make an
analogous interaction in CDK2, hence all effective inhibitors in the SAR series are
carboxylates. Furthermore, adding to the importance of the carboxylate for inhibitor binding
is the sensitivity for carboxylate stereochemistry, where the S-enantiomer is favoured over
16

ACCEPTED MANUSCRIPT
the R-enantiomer. The exception to this pattern is found for the most lipophilic carboxylates
(compounds 9 and 10) where both enantiomers bind equally well.
One distinguishing characteristic of CK2 is its ability to use GTP as the phosphate donor,[28]
a feature shared with calcium/calmodulin-dependent protein kinase II,[29] mst3 kinase,[30]
protein kinase C[31] and relatively few other protein kinases.[32] The structural mechanism
by which this occurs in CK2 is that the guanine base, which has a hydrogen bond acceptor–
donor–donor pattern (in contrast to adenine's donor–acceptor–empty pattern) binds to the
protein kinase hinge slightly shifted in order to align the two acceptor positions. The shift
creates a cavity, filled by a water molecule to match the adenine donor position. This
anomalous property may be related to the anomalously strong CK2 inhibition of the quinoline
derivatives.
(Insert Figure 7.)
3.4 SAR of Aurora A and B inhibition
A significant feature apparent from Table 1 is the inhibition of Aurora B by the quinoline-
derived inhibitors 13, 14, 15, and 16. In contrast, Aurora A is not significantly inhibited by
any of these inhibitors, despite its close similarity to Aurora B. This fact may provide key
information to identify the mechanisms of selectivity that are operative for the quinoline-
derived inhibitors.
The general insensitivity to the position of the hydroxyl substituent suggests that the
quinoline inhibitors bind to Aurora B without hinge binding to the hydroxyl group, unlike
either of the CDK2 co-crystal structures described above. However, the precise position of
the hydroxyl group does affect the sensitivity of Aurora B inhibition to changes in
stereochemistry and the relative orientation of the carboxylate group relative to the thiazole
17

ACCEPTED MANUSCRIPT
ring. For the benzoimidazole compounds (1 vs 2), there is strong sensitivity to
stereochemistry. For the 6-hydroxy quinoline acids (compounds 13 and 14), the sensitivity is
weaker, and for the 8-hydroxy quinoline acids (compounds 15 and 16) it is absent entirely.
These fundamental differences in SAR between CDK2/Cyclin A and the Aurora kinases are
not reflected by any obvious differences in the active sites of the proteins, and experimental
determination of binding modes in the Aurora kinases seems necessary to understand the
origins of the SAR observed for the Aurora kinases.
3.6 Pairwise SAR effects on strongest inhibitors
The exercise of explaining observed inhibitory modulation via pairwise comparisons that
focus on individual binding features provides both a test and a summary of SAR hypotheses.
3.6.1. Stereochemistry of the carboxylic acid substituent
As described above, the observed and mostly likely interaction of the carboxylate substituent
in the luciferin derivatives is the active site lysine (see also Figure 7). If this interaction exists
in strong inhibitors, as seems likely, changing the stereochemistry would either destroy the
interaction and weaken the inhibition, or the interaction would be preserved by a combination
of inhibitor rotation, thiazole rotation, and side chain flexibility. A more constrained ATP site
and/or binding mode would thus be more sensitive to the stereochemistry change. Four pairs
of inhibitors may be compared to test for these effects. Changing the stereochemistry of D-
luciferin (1) produces compound 2, which greatly weakens binding to most kinases strongly
inhibited by D-luciferin, however the effect on the DYRK-family is small. The
stereochemistry shift between 9 and 10 does not change the strong inhibition of CK2 but
stereochemical preference for the DYRK-family changes and now favours the R-enantiomer.
The corresponding changes between the quinoline enantiomeric pairs of 13/14 and 15/16
18

ACCEPTED MANUSCRIPT
show generally similar effects on DYRK 2, DYRK3, CK2, and Aurora B, with the S-
configuration giving the strongest inhibition.
3.6.2. Neutralization of the carboxylic acid substituent by esterification
If an anionic charge at carboxylate position is important (as we assume for active site lysine
salt bridge formation), its neutralization via methylation will show weakened binding. Indeed,
this is the case for the two pairs that involve strong inhibition of DYRK2 and DYRK3: 1/3
and 2/4. DYRK1a is unaffected by this change. However, for the methoxy-compounds 5/7
and 6/8, the situation is reversed and the more lipophilic ester derivatives are more active
than the free acids.
3.6.3. Modification of lipophilic bulk
The addition of methyl groups to the dihydrothiazole ring are seen with pairs 1/11, 2/12, 5/9,
and 6/10. The addition of bulk to luciferin (1/11) is fully tolerated by the strongly inhibited
kinases in the DYRK-family. The other strongly inhibited kinases loose activity upon
methylation of the dihydrothiazole ring. The enantiomeric 2/12 with R-configuration at the
carboxylate pair shows that methylation even increases inhibition of the DYRK-kinases. For
the methoxylated pairs 5/9 and 6/10 differing only by the stereochemistry of the carboxylate
on the dihydrothiazole, methylation is tolerated by the DYRK-kinases while the CK2-kinase
is substantially more inhibited.
3.6.4. Modifications of the benzothiazole or quinoline phenol
Methylation of the benzothiazole phenol eliminates strong binding when compared to the
phenolic benzothiazoles, except for 9 and 10 inhibition of CK2 and DYRK2. This is
consistent with most binding modes involving hydrogen bonding of the alcohol to the hinge.
Unique characteristics of CK2 and DYRK2 have been discussed above and include GTP
19

ACCEPTED MANUSCRIPT
binding to CK2 and the generally strong inhibition of DYRK2 by all compounds. The
methoxy derivatives require a carboxylic acid for CK2 inhibition. In striking contrast to
benzothiazoles, the quinolines inhibit only Aurora B, CK2 and DYRK2 strongly, as discussed
above. Binding patterns for these three proteins are qualitatively similar.
4. Conclusion
The observation that D-luciferin inhibits quite selectively several protein kinases in the
CMGC-group has two important consequences. First, D-luciferin and cognates represent a
useful fragment-like scaffold for protein kinase inhibitor design with an interesting intrinsic
selectivity for the DYRK-family of protein kinases complementing the already known
molecular scaffolds. Furthermore, the profiling and SAR analysis here also demonstrate good
potential for achieving a wide range of selectivity patterns among the other drug target
kinases described here. With multiple binding geometries at the hinge, and multiple positions
for further derivatisation, the scaffolds provide myriad possibilities for drug design. On the
other hand, the multiple binding geometries limits the extension of the qualitative
understanding of the SAR derived from the structures of CDK2 complexes to other kinases.
Second, the potential for ligand interactions to be shared between protein kinases and
luciferases deserves close attention, as luciferin/luciferase based assays are in widespread
use. Awareness of this possibility may aid interpretation of apparently anomalous results.
5. Experimental
5.1 Protein Expression, Purification and Crystallization
20

ACCEPTED MANUSCRIPT
CDK2 was expressed with an N-terminal His-tag in BL21(DE3) cells. The protein was
purified via a His-tag column (50 mM sodium phosphate, 500 mM NaCl, pH 7.0) and eluted
via an imidazole gradient (50 mM sodium phosphate, 300 mM NaCl, 500 mM imidazole pH
7.0). The combined fractions with the CDK2 were loaded on a S200 16/60 gelfiltration
column (10 mM Hepes, 20 mM NaCl, 2 mM DTT, pH 7.4). CDK2 formed crystals
spontaneously after concentrating the fractions from the gel filtration above 6 mg/ml at 4 °C.
5.2 Structure determination
Crystals were soaked with either D-luciferin or the derivatives and cryoprotected with
glycerol prior to the freezing. The crystals were measured at the Berlin Electron Storage Ring
Society for Synchrotron Radiation (BESSY II) in Berlin/Germany.
5.3 Thermal shift assay
20 µl of CDK2 wild-type protein 0.7-0.8 mg/ml (50 mM phosphate buffer pH 7.5, 150 mM
NaCl) was mixed with 5 µL 6% SyproOrange (Sigma). 1 µL of DMSO or 1µL of 10 mM
inhibitor dissolved in DMSO was added. The thermal shift assay was done on a Miniopticom
(Biorad) by heating the sample slowly from 20 °C to 90 °C with a step interval of 1/3 °C. All
experiments were performed in triplicate.
5.4 Preparation of inhibitors
The inhibitors were prepared by a condensation between cysteine and a benzothiazole nitrile
or a quinoline nitrile derivative. Demethylation of the aromatic methoxy group or
1
esterification of the dihydrothiazolecarboxylate group was preformed as needed. H-NMR
and ESI mass spectral data for all compounds were in accordance with the literature and are
given in the Supplementary Information.
5.5 Determination of protein kinase inhibition
21

ACCEPTED MANUSCRIPT
The determination of protein kinase inhibition was performed at the International Centre for
Kinase Profiling at the University of Dundee, U.K. The method used is a radioactive filter
binding assay using 33P ATP.[3, 33] The ATP concentrations were at or below the calculated
Km for ATP for each particular kinase.
ACKNOWLEDGEMENT
Provision of beam time at Bessy II, Berlin Germany at BL14.1 is gratefully acknowledged.
Special thanks to the Helmholtz center, for supporting the travel expenses for Ulli
Rothweiler.
REFERENCES
[1] A. Lundin, J. Eriksson, A real-time bioluminescent HTS method for measuring protein
kinase activity influenced neither by ATP concentration nor by luciferase inhibition, Assay
and drug development technologies, 6 (2008) 531-541.
[2] P. Singh, B.J. Harden, B.J. Lillywhite, P.M. Broad, Identification of kinase inhibitors by
an ATP depletion method, Assay and drug development technologies, 2 (2004) 161-169.
[3] J. Bain, L. Plater, M. Elliott, N. Shapiro, C.J. Hastie, H. McLauchlan, I. Klevernic, J.S.
Arthur, D.R. Alessi, P. Cohen, The selectivity of protein kinase inhibitors: a further update,
The Biochemical journal, 408 (2007) 297-315.
[4] E.H. White, F. McCapra, G.F. Field, The structure and synthesis of firefly luciferin, J.
Am. Chem. Soc., 85 (1963) 337-343.
[5] D.C. McCutcheon, M.A. Paley, R.C. Steinhardt, J.A. Prescher, Expedient synthesis of
electronically modified luciferins for bioluminescence imaging, J Am Chem Soc, 134 (2012)
7604-7607.
[6] G. Meroni, M. Rajabi, E. Santaniello, D-Luciferin, derivatives and analogues: synthesis
and in vitro/in vivo luciferase-catalyzed bioluminicent activity, ARKIVOC, 2009 (2009) 265-
288.
[7] H.-J. Böhm, S. Brode, U. Hesse, G. Klebe, Oxygen and Nitrogen in Competitive
Situations: Which is the Hydrogen-Bond Acceptor?, Chemistry – A European Journal, 2
(1996) 1509-1513.
[8] P.D. Jeffrey, A.A. Russo, K. Polyak, E. Gibbs, J. Hurwitz, J. Massague, N.P. Pavletich,
Mechanism of CDK activation revealed by the structure of a cyclinA-CDK2 complex,
Nature, 376 (1995) 313-320.
[9] A.A. Russo, P.D. Jeffrey, A.K. Patten, J. Massague, N.P. Pavletich, Crystal structure of
the p27Kip1 cyclin-dependent-kinase inhibitor bound to the cyclin A-Cdk2 complex, Nature,
382 (1996) 325-331.
22

ACCEPTED MANUSCRIPT
[10] R.A. Engh, D. Bossemeyer, The protein kinase activity modulation sites: mechanisms
for cellular regulation - targets for therapeutic intervention, Advances in enzyme regulation,
41 (2001) 121-149.
[11] B. Smith, F. Medda, V. Gokhale, T. Dunckley, C. Hulme, Recent advances in the design,
synthesis, and biological evaluation of selective DYRK1A inhibitors: a new avenue for a
disease modifying treatment of Alzheimer's?, ACS chemical neuroscience, 3 (2012) 857-872.
[12] M.A. Beniddir, E. Le Borgne, B.I. Iorga, N. Loaec, O. Lozach, L. Meijer, K. Awang, M.
Litaudon, Acridone alkaloids from Glycosmis chlorosperma as DYRK1A inhibitors, Journal
of natural products, 77 (2014) 1117-1122.
[13] J. Bain, H. McLauchlan, M. Elliott, P. Cohen, The specificities of protein kinase
inhibitors: an update, The Biochemical journal, 371 (2003) 199-204.
[14] M. Gompel, M. Leost, E.B. De Kier Joffe, L. Puricelli, L.H. Franco, J. Palermo, L.
Meijer, Meridianins, a new family of protein kinase inhibitors isolated from the ascidian
Aplidium meridianum, Bioorganic & medicinal chemistry letters, 14 (2004) 1703-1707.
[15] C. Neagoie, E. Vedrenne, F. Buron, J.Y. Merour, S. Rosca, S. Bourg, O. Lozach, L.
Meijer, B. Baldeyrou, A. Lansiaux, S. Routier, Synthesis of chromeno[3,4-b]indoles as
Lamellarin D analogues: a novel DYRK1A inhibitor class, European journal of medicinal
chemistry, 49 (2012) 379-396.
[16] M. Debdab, F. Carreaux, S. Renault, M. Soundararajan, O. Fedorov, P. Filippakopoulos,
O. Lozach, L. Babault, T. Tahtouh, B. Baratte, Y. Ogawa, M. Hagiwara, A. Eisenreich, U.
Rauch, S. Knapp, L. Meijer, J.P. Bazureau, Leucettines, a class of potent inhibitors of cdc2-
like kinases and dual specificity, tyrosine phosphorylation regulated kinases derived from the
marine sponge leucettamine B: modulation of alternative pre-RNA splicing, Journal of
medicinal chemistry, 54 (2011) 4172-4186.
[17] P. Kassis, J. Brzeszcz, V. Beneteau, O. Lozach, L. Meijer, R. Le Guevel, C. Guillouzo,
K. Lewinski, S. Bourg, L. Colliandre, S. Routier, J.Y. Merour, Synthesis and biological
evaluation of new 3-(6-hydroxyindol-2-yl)-5-(Phenyl) pyridine or pyrazine V-Shaped
molecules as kinase inhibitors and cytotoxic agents, European journal of medicinal
chemistry, 46 (2011) 5416-5434.
[18] Y. Loidreau, P. Marchand, C. Dubouilh-Benard, M.R. Nourrisson, M. Duflos, O.
Lozach, N. Loaec, L. Meijer, T. Besson, Synthesis and biological evaluation of N-
arylbenzo[b]thieno[3,2-d]pyrimidin-4-amines and their pyrido and pyrazino analogues as
Ser/Thr kinase inhibitors, European journal of medicinal chemistry, 58 (2012) 171-183.
[19] Y. Loidreau, P. Marchand, C. Dubouilh-Benard, M.R. Nourrisson, M. Duflos, N. Loaec,
L. Meijer, T. Besson, Synthesis and biological evaluation of N-aryl-7-
methoxybenzo[b]furo[3,2-d]pyrimidin-4-amines and their N-arylbenzo[b]thieno[3,2-
d]pyrimidin-4-amine analogues as dual inhibitors of CLK1 and DYRK1A kinases, European
journal of medicinal chemistry, 59 (2013) 283-295.
[20] L. Demange, F.N. Abdellah, O. Lozach, Y. Ferandin, N. Gresh, L. Meijer, H. Galons,
Potent inhibitors of CDK5 derived from roscovitine: synthesis, biological evaluation and
molecular modelling, Bioorganic & medicinal chemistry letters, 23 (2013) 125-131.
[21] E. Deau, Y. Loidreau, P. Marchand, M.R. Nourrisson, N. Loaec, L. Meijer, V. Levacher,
T. Besson, Synthesis of novel 7-substituted pyrido[2',3':4,5]furo[3,2-d]pyrimidin-4-amines
and their N-aryl analogues and evaluation of their inhibitory activity against Ser/Thr kinases,
Bioorganic & medicinal chemistry letters, 23 (2013) 6784-6788.
[22] O. Dehbi, A. Tikad, S. Bourg, P. Bonnet, O. Lozach, L. Meijer, M. Aadil, M. Akssira, G.
Guillaumet, S. Routier, Synthesis and optimization of an original V-shaped collection of 4-7-
disubstituted pyrido[3,2-d]pyrimidines as CDK5 and DYRK1A inhibitors, European journal
of medicinal chemistry, 80 (2014) 352-363.
23

ACCEPTED MANUSCRIPT
[23] Y. Ogawa, Y. Nonaka, T. Goto, E. Ohnishi, T. Hiramatsu, I. Kii, M. Yoshida, T. Ikura,
H. Onogi, H. Shibuya, T. Hosoya, N. Ito, M. Hagiwara, Development of a novel selective
inhibitor of the Down syndrome-related kinase Dyrk1A, Nature communications, 1 (2010)
86.
[24] K. Anderson, Y. Chen, Z. Chen, R. Dominique, K. Glenn, Y. He, C. Janson, K.C. Luk,
C. Lukacs, A. Polonskaia, Q. Qiao, A. Railkar, P. Rossman, H. Sun, Q. Xiang, M. Vilenchik,
P. Wovkulich, X. Zhang, Pyrido[2,3-d]pyrimidines: discovery and preliminary SAR of a
novel series of DYRK1B and DYRK1A inhibitors, Bioorganic & medicinal chemistry letters,
23 (2013) 6610-6615.
[25] M. Soundararajan, A.K. Roos, P. Savitsky, P. Filippakopoulos, A.N. Kettenbach, J.V.
Olsen, S.A. Gerber, J. Eswaran, S. Knapp, J.M. Elkins, Structures of Down Syndrome
Kinases, DYRKs, Reveal Mechanisms of Kinase Activation and Substrate Recognition,
Structure, 21 (2013) 986-996.
[26] T. Tahtouh, J.M. Elkins, P. Filippakopoulos, M. Soundararajan, G. Burgy, E. Durieu, C.
Cochet, R.S. Schmid, D.C. Lo, F. Delhommel, A.E. Oberholzer, L.H. Pearl, F. Carreaux, J.P.
Bazureau, S. Knapp, L. Meijer, Selectivity, cocrystal structures, and neuroprotective
properties of leucettines, a family of protein kinase inhibitors derived from the marine sponge
alkaloid leucettamine B, Journal of medicinal chemistry, 55 (2012) 9312-9330.
[27] G. Cozza, L.A. Pinna, S. Moro, Protein kinase CK2 inhibitors: a patent review, Expert
Opin Ther Pat, 22 (2012) 1081-1097.
[28] K. Niefind, M. Putter, B. Guerra, O.G. Issinger, D. Schomburg, GTP plus water mimic
ATP in the active site of protein kinase CK2, Nature structural biology, 6 (1999) 1100-1103.
[29] S.L. Bostrom, J. Dore, L.C. Griffith, CaMKII uses GTP as a phosphate donor for both
substrate and autophosphorylation, Biochemical and biophysical research communications,
390 (2009) 1154-1159.
[30] K. Schinkmann, J. Blenis, Cloning and characterization of a human STE20-like protein
kinase with unusual cofactor requirements, The Journal of biological chemistry, 272 (1997)
28695-28703.
[31] M. Gschwendt, W. Kittstein, K. Kielbassa, F. Marks, Protein-Kinase C-Delta Accepts
Gtp for Autophosphorylation, Biochemical and biophysical research communications, 206
(1995) 614-620.
[32] C.A. Smith, M. Toth, H. Frase, L.J. Byrnes, S.B. Vakulenko, Aminoglycoside 2''-
phosphotransferase IIIa (APH(2'')-IIIa) prefers GTP over ATP: structural templates for
nucleotide recognition in the bacterial aminoglycoside-2'' kinases, The Journal of biological
chemistry, 287 (2012) 12893-12903.
[33] C.J. Hastie, H.J. McLauchlan, P. Cohen, Assay of protein kinases using radiolabeled
ATP: a protocol, Nature protocols, 1 (2006) 968-971.
24

ACCEPTED MANUSCRIPT
Captions to the Figures and Schemes.
Figure 1. SAR variations applied to the D-luciferin scaffold
Scheme 1. Outline of the preparation of the benzothiazole SAR-library. The quinoline library
was prepared by the same methods. Reagents and conditions: (a) Na2CO3 (aq), rt, 3h; (b)
SOCl2, Methanol, rt, 1h.
Figure 2. ATP binding pocket of CDK2. A and B: electron density of the inhibitors before
the refinement, and C and D: after the refinement. A and C represents the complex between
CDK2 and 1, whereas B and D represents the complexes with 15. The electron density is
shown in blue at 1 σ
fference density in green and red at 2.5 σ for A and B, and at 3 σ
for C and D.
Figure 3. Orientation of 1 (magenta) and 15 (orange) in the ATP binding pocket of CDK2.
Compound 15 is rotated approximately by 90 degrees relative to 1 and penetrates deeper into
the ATP pocket. The protein part of CDK2 in complex with 1 is represented in green; with 15
using yellow carbon atoms. The position of the OH group on the benzothiazole and quinoline
rings defines the orientation and binding mode of the inhibitor. Both inhibitors bind to the
hinge, but use different backbone carbonyl groups for the H-bond.
Figure 4. Binding pocket of CDK2 bound to compounds 1 (A) and 15 (B).
Figure 5: Superposition of CDK2/1 (green) and CDK2/15 (yellow). The CDK2/15 structure
has a secondary binding site for compound 15 in the C-lobe not present in the CDK2/1
structure.
Figure 6. DYRK1a kinases (3ANR, 3ANQ, cyan) have a differently shaped ATP site in
comparison with the CDK2/1 structure (yellow) of this study, due to conformation or rotamer
25

ACCEPTED MANUSCRIPT
differences in combination with differences in sequence. As a result, luciferin occupies a
volume distinct from that of the DYRK inhibitors (see text).
Figure 7. Carboxylate groups of CK2 inhibitors cluster as salt bridge partners to the active
site lysine.
26

ACCEPTED MANUSCRIPT
Ring size and geometry
Stereochemistry
O
N
S
Removal of acidic
proton by
esterification
N
OH
H
Hydrogen
bond donation
S
HO
Lipophilic bulk