2-Azo-, 2-diazocine-thiazols and 2-azo-imidazoles as photoswitchable kinase inhibitors: Limitations and pitfalls of the photoswitchable inhibitor approach

Article abstract of DOI:10.1039/c9pp00010k

In photopharmacology, photoswitchable compounds including azobenzene or other diarylazo moieties exhibit bioactivity against a target protein typically in the slender E-configuration, whereas the rather bulky Z-configuration usually is pharmacologically less potent. Herein we report the design, synthesis and photochemical/inhibitory characterization of new photoswitchable kinase inhibitors targeting p38α MAPK and CK1δ. A well characterized inhibitor scaffold was used to attach arylazo- and diazocine moieties. When the isolated isomers, or the photostationary state (PSS) of isomers, were tested in commonly used in vitro kinase assays, however, only small differences in activity were observed. X-ray analyses of ligand-bound p38α MAPK and CK1δ complexes revealed dynamic conformational adaptations of the protein with respect to both isomers. More importantly, irreversible reduction of the azo group to the corresponding hydrazine was observed. Independent experiments revealed that reducing agents such as DTT (dithiothreitol) and GSH (glutathione) that are typically used for protein stabilization in biological assays were responsible. Two further sources of error are the concentration dependence of the E-Z-switching efficiency and artefacts due to incomplete exclusion of light during testing. Our findings may also apply to a number of previously investigated azobenzene-based photoswitchable inhibitors.

Full text of DOI:10.1039/c9pp00010k

Photochemical &
Photobiological Sciences
View Article Online
View Journal
PAPER
2-Azo-, 2-diazocine-thiazols and 2-azo-
imidazoles as photoswitchable kinase inhibitors:
limitations and pitfalls of the photoswitchable
inhibitor approach†
Cite this: DOI: 10.1039/c9pp00010k
c
Miriam Schehr,a Chiara Ianes,b Jörn Weisner,c Linda Heintze,d Matthias P. Müller,
Christian Pichlo,e Julia Charl,e Elena Brunstein,e Julia Ewert,a Marc Lehr,a
Ulrich Baumann,e Daniel Rauh, c Uwe Knippschild,b Christian Peifer *d and
Rainer Herges*a
In photopharmacology, photoswitchable compounds including azobenzene or other diarylazo moieties
exhibit bioactivity against a target protein typically in the slender E-configuration, whereas the rather
bulky Z-configuration usually is pharmacologically less potent. Herein we report the design, synthesis and
photochemical/inhibitory characterization of new photoswitchable kinase inhibitors targeting p38α MAPK
and CK1δ. A well characterized inhibitor scaffold was used to attach arylazo- and diazocine moieties.
When the isolated isomers, or the photostationary state (PSS) of isomers, were tested in commonly used
in vitro kinase assays, however, only small differences in activity were observed. X-ray analyses of ligand-
bound p38α MAPK and CK1δ complexes revealed dynamic conformational adaptations of the protein with
respect to both isomers. More importantly, irreversible reduction of the azo group to the corresponding
hydrazine was observed. Independent experiments revealed that reducing agents such as DTT (dithio-
threitol) and GSH (glutathione) that are typically used for protein stabilization in biological assays were
responsible. Two further sources of error are the concentration dependence of the E–Z-switching
efficiency and artefacts due to incomplete exclusion of light during testing. Our findings may also apply to
a number of previously investigated azobenzene-based photoswitchable inhibitors.
Received 7th January 2019,
Accepted 21st March 2019
DOI: 10.1039/c9pp00010k
rsc.li/pps
compounds, namely photopharmacology, is a rapidly evolving
field with great potential. The goal is to control the activity of
Introduction
Photoswitchable molecules which undergo conformational or molecules by changing their geometry and thus their bioactiv-
configurational changes triggered by light show great potential ity using light as a non-invasive trigger. This approach opens
in versatile applications such as molecular machines,1–3 con- the possibility of a spatially and temporally controlled drug
trolling conductivity,4,5 or pharmacology.6–9 Especially utiliz- therapy. Furthermore, fundamental biological processes can
ation of photoswitchable moieties in pharmacologically active be precisely studied using photoresponsive pharmacological
tools. Following this strategy, a number of photoswitchable
molecules for different targets and pharmacological purposes
aOtto Diels-Institute of Organic Chemistry, Christian Albrechts University Kiel,
Otto-Hahn-Platz 4, 24118 Kiel, Germany. E-mail: rherges@oc.uni-kiel.de
bDepartment of General and Visceral Surgery, Ulm University Hospital,
Albert-Einstein-Allee 23, 89081 Ulm, Germany
cFaculty of Chemistry and Chemical Biology, TU Dortmund University and Drug
Discovery Hub Dortmund (DDHD) am Zentrum für integrierte Wirkstoffforschung
(ZIW), Otto-Hahn-Straße 4a, 44227 Dortmund, Germany
dInstitute of Pharmacy, Christian Albrechts University Kiel, Gutenbergstraße 76,
24118 Kiel, Germany. E-mail: cpeifer@pharmazie.uni-kiel.de
eDepartment of Chemistry, University of Cologne, Otto-Fischer-Straße 12-14,
50674 Cologne, Germany
have been developed in academia. A selection of examples is
presented in Table 1. Mostly, azobenzene moieties were used
as the E/Z-photoswitchable units allowing for rather large
steric changes of the respective pharmacophore. Thus, for
smartly designed and optimized inhibitors one could ideally
expect one isomer to be biologically highly active against a
specific target, whereas the other isomer should show no
activity at all. At least, both isomers should exhibit signifi-
cantly different activities. Typically, for azobenzenes, the
E-isomer has a higher biological activity in comparison to the
Z-isomer. However, frequently the difference in biological
activity is rather small. Examples of efficient azobenzene based
†Electronic supplementary information (ESI) available. CCDC 1885848. For ESI
and crystallographic data in CIF or other electronic format see DOI: 10.1039/
c9pp00010k
This journal is © The Royal Society of Chemistry and Owner Societies 2019
Photochem. Photobiol. Sci.

View Article Online
Paper
Photochemical & Photobiological Sciences
Table 1 IC50 values of the different isoforms and the corresponding factor of selected photoswitchable biomolecules published by different
groups
Study
Selected target
Photoswitchable moiety IC50 value (E/trans or open) IC50 value (Z/cis or closed) Ratio
Morton et al.14
König et al.11
König et al.15
Feringa et al.16
Feringa et al.10
Fuchter et al.17
König et al.18
König et al.19
Grøtli et al.12
Branda et al.20
Cysteine protease calpain
Human carbonic anhydrase I Diarylethene
Azobenzene
45 nM
8 nM
4.2 µM
41 nM
21.7 µM
2.5 µM
18 µM
29.6 µM
150 nM
580 nM
65 nM (KD value)
175 nM
400 nM
92.3 µM
91 nM
0.56 µM
0.18 µM
4.5 M
19.9 µM
580 nM
inactive
146 nM (KD value)
3.8-fold
50-fold
22-fold
2-fold
39-fold
14-fold
4-fold
1.5-fold
4-fold
—
Histone deacetylase
Proteasome
Histone deacetylase
Amidohydrolase
Histone deacetylase
Histone deacetylase
RET kinase
Diarylmaleimide
Azobenzene
Azobenzene
Azobenzene
Dithienylethene
Fulgimides
Azobenzene
Diarylethene
Azobenzene
Protein kinase C
Liskamp et al.13 Spleen tyrosine kinase
11-fold
photoswitchable inhibitors include
a
histone deacetylase
studied by Feringa et al.10 where a 39-fold difference between
the E- and the Z-isomer was observed. Comparably, König
et al.11 were able to achieve a 50-fold difference in activity of a
carbonic anhydrase using the open or closed form of a diaryl-
ethene as a photoswitch. Regarding various targets of photo-
switchable inhibitors, only few protein kinase inhibitors based
on azobenzenes were reported, including examples of Grøtli
et al.12 (E/Z: 4-fold), and Liskamp et al.13 (E/Z: 11-fold).
Protein kinases are phosphotransferases that catalyse the
transfer of the γ-phosphate group from ATP to a hydroxyl
group of substrates. They play important roles in signal trans-
duction cascades regulating different cellular processes like
cell differentiation, cell proliferation or apoptosis. Thus, dysre-
gulation of key protein kinases often leads to severe disorders
like cancer, inflammation, or neurodegenerative diseases.21–25
The human kinome consists of more than 500 different
protein kinases, rendering selectivity of an inhibitor for a
specific kinase a key issue. Until today, more than 40 approved
protein kinase inhibitors are used in therapy, mostly for
cancer treatment.26–28 In this context, photoswitchable kinase
inhibitors allowing spatial and temporal control by light would
offer great therapeutical advantages. Such an inhibitor could
be specifically activated on a focussed site of action, which
may lead to high efficacy and reduced systemic side effects.
Motivated by the relatively few examples of photoswitchable
protein kinase inhibitors reported so far, we started to investi-
gate the applicability of photoswitches towards this compound
class in more detail.
In the present study we report on the design, synthesis and
biological evaluation of photoswitchable inhibitors for p38α
Mitogen-activated Protein Kinase (p38α MAPK) and the highly
related Casein Kinase 1δ (CK1δ). As model compound we used
the known 4,5-diarylimidazole compound 1 which is a highly
potent inhibitor of both p38α MAPK and CK1δ.29 Structural
data for 1 in the ATP binding pocket and detailed structure–
activity relationships (SAR) for CK1δ were available and pro-
Fig. 1 Top: 2D ligand-interaction diagram (LID) and corresponding 3D
binding mode showing the protein surface of original ligand 1 (tur-
quoise) in the ATP active site of p38α MAPK as determined by X-ray crys-
tallography (pdb 5ML5).29 Bottom: LID and 3D binding mode of
designed photoswitchable azo-compound 2 in the ATP active site of
p38α MAPK based on docking calculations. The ligand reveals a plausible
binding mode only in the E-configuration, whereas the Z-configured
ligand did not adequately fit into the ATP pocket due to significant steric
clashes. Similar results were obtained for designed ligand 3 in p38α
MAPK, and for 2 and 3 in the highly related structure of CK1δ (pdb
5MQV, not shown). For details see ESI.†
vided the basis for a rational design strategy of the photo- unit. The heterocyclic, aromatic five-ring was included as part
switchable inhibitors (Fig. 1).29 Preliminary docking studies of the switching diarylazo system (Fig. 1). Azobenzene-type
revealed the methylthioether substituted 2-imidazole position photoswitches with one of the benzene rings replaced by a
in 1 as the most suitable position to attach the switchable azo 5-membered heterocyclic ring often exhibit favourable photo-
Photochem. Photobiol. Sci.
This journal is © The Royal Society of Chemistry and Owner Societies 2019

View Article Online
Photochemical & Photobiological Sciences
Paper
physical properties,30 however, Z-1H-azoimidazoles rapidly iso- the thermodynamically stable Z-configuration of 4.
merize back to the more stable Z-isomer via tautomeric struc- Consequently,
a
higher bioactivity of the metastable
tures.31 Therefore, the imidazole unit must be N-methylated, E-configuration is expected.
or the imidazole has to be replaced by e.g. a thioazole ring
(Fig. 1, bottom) to obtain a system with a sufficient half-life in
the Z-configuration.
Results and discussion
Synthesis and photophysical properties
According to the crystallographically determined binding
mode of 1 in p38α MAPK,29 this moiety is situated in a narrow
Based on the design concept of the photoswitchable p38α
channel formed by Tyr35, Val38, Lys53, Glu71 and Phe169
(amongst further residues) in close proximity to the crucial
activation loop and the glycine-rich loop. Therefore, this
pocket of the protein kinase should be able to accommodate
the azoaryl moiety in E-configuration, but not in
Z-configuration, thus translating into different photorespon-
sive bioactivities, respectively. Furthermore, we extended this
concept by a thiodiazocine moiety attached to the thiazole core
in a similar manner (Fig. 2). The diazocine moiety was chosen
because this class of compounds has photochemical pro-
perties that are highly favorable for photoswitchable ligand
design. Compared to azobenzenes, diazocines are thermo-
dynamically more stable in the Z-configuration and can be
switched to the metastable E-configuration upon irradiation
(see Scheme 1).32
MAPK inhibitor outlined above, we aimed at the synthesis and
photochemical/photophysical characterization of target com-
pounds 2–4 (summarized in Scheme 2).
Our straightforward seven-step synthesis of 2-azo-thiazole
inhibitor 2 is presented in Scheme 3 (for detailed synthetic
procedures see ESI†). Starting material 2-amino-4-methyl-
pyridine 5 was successively double Boc-protected with di-tert-
butyl dicarbonate at first in tert-butanol and then using 4-di-
methylaminopyridine (DMAP) in THF at room temperature.33
Attempts of direct double protection always led to a mixture of
single and double Boc-protected products 6 and 7. The double
protection, however, is necessary for the synthesis of the etha-
none derivative 9. Here, the methyl group was deprotonated
with sodium hexamethyldisilazane (NaHMDS) in THF and
reacted with ethyl-4-fluorobenzoate 8 to obtain the ethanone
derivative 9.34 During this reaction, one of the Boc-groups is
cleaved. Using the single Boc-protected methylpyridine 6 for
the formation of ethanone 9 led to complete decomposition of
the starting material 6. After NBS bromination of the ethanone
in the α-position, the obtained α-bromoketone 10 was reacted
with thiosemicarbazide 11 to form the thiazole ring functiona-
lized with a hydrazine moiety. The hydrazine could be selec-
tively oxidized in situ by bubbling compressed air through the
In docking studies, the metastable E-configuration of 4 was
calculated to fit more precisely into the binding pocket than
Fig. 2 Top: Calculated 2D ligand-interaction diagram (LID) and 3D
binding mode showing the protein surface of designed ligand
4
)
(E-configuration) within the ATP active site of p38α MAPK (pdb 5ML529
determined by docking. Bottom: Distorted suboptimal binding pose of
compound 4 in the Z-configuration causing sterical clashes (dotted
orange lines). For details see ESI.†
Scheme 2 Potential photoswitchable target inhibitors 2–4 derived
from the known CK1δ/p38α MAPK inhibitor 1.29 In target structure 2 and
3 the 1H-imidazole is replaced by 1,3-thiazole or 1-methyl imidazole
and the thiomethyl substituent is replaced by an azophenyl group. In
target inhibitor 4 a photoswitchable S-diazocine unit is attached to the
2-position of the 1,3-thiazole.
Scheme 1 The diazocine moiety is thermodynamically more stable in
the Z-configuration and can be converted to the metastable
E-configuration upon irradiation with 405 nm.
This journal is © The Royal Society of Chemistry and Owner Societies 2019
Photochem. Photobiol. Sci.

View Article Online
Paper
Photochemical & Photobiological Sciences
prepared according to literature.38,39 In the last synthesis step
the pyridine moiety functionalized via peptide coupling with
3-(2,5-dimethoxy-phenyl)propionic acid 14 was introduced in
the azoimidazole 19 using palladium catalyzed conditions to
afford target compound 3.40
To obtain the 2-diazocine-thiazole 4, the synthetic route
was started with the synthesis of the functionalized diazocine
precursor 26 (Scheme 5).
Disulfide 23 was reduced with sodium borohydride and
then reacted in situ with the functionalized benzyl bromide 22
to afford thioether 24.41 The carboxylic acid was converted to
the acid amide 25, which could further be functionalized with
Lawesson’s reagent to yield thioamide 26.42 In the next step,
the thioamide functionalized diazocine precursor 26 was used
in a ring closing reaction with α-bromoketone 10 forming the
thiazole compound 27. After peptide coupling with 3-(2,5-
dimethoxyphenyl)propionic acid 14 followed by reductive azo
condensation with lead, a small portion of the desired diazo-
cine 4 could be isolated directly.43 The majority of the product
Scheme 3 Reaction conditions: (i) Boc anhydride, t-butanol, 30 °C,
68%; (ii) Boc anhydride, THF, 99%; (iii) NaHMDS, THF, 0 °C, 60%; (iv) NBS,
AcOH/DCM, 97%; (v) AcOH, compressed air, THF, 80 °C, 42%; (vi) HCl,
ethyl acetate, 55 °C, 97%; (vii) T3P® (propylphosphonic anhydride),
DIPEA, 95 °C, 48%.
however was
a mixture of azoxy compounds 29a/29b
(Scheme 6). The latter mixture was converted to the diazocine
4 without further purification by another two-step procedure.
The non-defined mixture of 29a/29b was reduced with lead
powder using high power ultrasound conditions to yield hydra-
zine 30. Hydrazine 30 was then oxidized by copper(II)chloride
in a methanolic sodium hydroxide solution to again give the
target diazocine 4.
solution yielding the azobenzene moiety 12.35 Deprotection of
the 2-aminopyridine 12 with hydrochloric acid led to the
amine 13, which was functionalized with the 3-(2,5-dimethoxy-
phenyl)propionic acid 14 via peptide coupling conditions to
afford target compound 2.36
Using these combined techniques, the 2-diazocine-thiazol 4
could be obtained with a total yield of 12%.
The 2-azo-imidazole inhibitor 3 was obtained by a four-step
synthesis (Scheme 4). In the first synthetic step, the 4-fluoro-
phenyl moiety 17 was introduced via Suzuki cross coupling of
4-fluorophenylboronic acid 16 and 4-bromo-1-methyl-1H-imid-
azole 15.37 In the next step, to afford the 2-azoimidazole 19,
the imidazole 2-position was lithiated with n-butyllithium and
reacted with 4-fluorotetrafluorodiazonium salt 18, which was
Scheme 5 Reaction conditions: (i) NaBH4, THF, 60 °C, 35%; (ii) 1. Oxalyl
chloride, 2. aq. ammonia, ethyl acetate, quant. (iii) Lawessońs reagent,
THF, 80 °C, 78%; (iv) 1. MeCN, 2. AcOH/HCl, 80 °C, 36%; (v) T3P®,
DIPEA, 95 °C, ethyl acetate, 37%; (vi) 1. Pb, NEt3/formic acid/MeOH/H2O,
2. Cu(I)Cl, NaOH in MeOH, compressed air, 12%.
Scheme 4 Reaction conditions: (i) Pd[dppf]Cl2, BnNEt3Cl, CsF,
toluene/water, 90 °C, 84%; (ii) n-BuLi, THF, −78 °C, 19%; (iii) T3P®,
DIPEA, ethyl acetate, 95 °C, 57%; (iv) Pd(OAc)2, K2CO3, DMAc, 150 °C,
18%.
Photochem. Photobiol. Sci.
This journal is © The Royal Society of Chemistry and Owner Societies 2019

View Article Online
Photochemical & Photobiological Sciences
Paper
291 nm and λmax = 408 nm (ππ*) with a slight shoulder at
450 nm were observed (black line). Irradiation of this species
with 420 nm gives the PSS containing both, E- and
Z-configurations (red line) with absorption maxima at λmax
=
290 nm, λmax = 375 nm and λmax = 450 nm. Again, reverse iso-
merization can also be initiated by irradiation of the nπ* band
with light of 525 nm and, similar to compound 2, the back iso-
Scheme 6 Reaction conditions: (i) Pb, NEt3/formic acid/MeOH/H2O, merization cannot be achieved quantitatively (Fig. 3, top right,
high power ultrasound; (ii) Cu(II)Cl2, NaOH in MeOH, compressed air.
blue line). However, both compounds 2 and 3 show excellent
photostability in DMSO when repeatedly photoswitched between
the E- and Z-configuration, respectively (Fig. 3, bottom).
The photochemical properties of key compounds 2–4 were
investigated using UV/vis and 1H NMR spectroscopy.
Unfortunately, besides poor solubility in water, compounds
2–4 showed in aqueous medium very poor photochemical pro-
perties. Since an adequate photochemical characterization was
not possible in aqueous solutions, we performed the photo-
chemical analysis in DMSO solution. In Fig. 3, UV/vis spectra
of the 2-azo-thiazole 2 and -imidazole 3 including photostabil-
ity tests are shown. The 2-azo-thiazole 2 in the E-configuration
(black line) displays absorption maxima at λmax = 294 nm and
λmax = 417 nm. After irradiation with 435 nm the ππ* band
decreases significantly until the photostationary state (PSS)
between the E- and the Z-isomer (red line) is reached, with
absorption maxima at 289 nm and 376 nm, respectively. To
trigger the reverse isomerisation (Z towards E), irradiation into
the nπ* band of the Z-isomer with 525 nm was used (Fig. 3,
top left, blue line). Due to an overlap of the ππ* band of the
E-isomer and the nπ* band of the Z-isomer, back isomerization
could not be achieved quantitatively (Fig. 3, blue line). The
UV/vis spectra of the 2-azo-imidazole 3 are similar to those of
The UV absorption of the diazocine moiety of 2-diazocine-
thiazole 4 is relatively weak compared to the UV absorption of
the residual molecule. Therefore, only the visible absorption of
the diazocine (420–700 nm) is displayed in Fig. 4 (for complete
UV/vis spectrum see ESI†). The UV spectrum of the Z-isomer of
4 is plotted in black. Upon irradiation with 405 nm the nπ*
band of the diazocine moiety in the E-configuration with λmax
= 520 nm emerges (Fig. 4, red line). Back isomerization can be
achieved using 525 nm quantitatively. Similar to the switch-
able inhibitors 2 and 3, compound 4 shows excellent photo-
stability in solution when subjected to repeated irradiation
with 405 and 525 nm in an alternating sequence (Fig. 4).
Photostationary states and half-lives (t1/2) of the respective
E/Z-isomers were measured using either 1H NMR or UV/vis
spectroscopy in deuterated DMSO. The results obtained for
compounds 2–4 are summarized in Table 2.
Although the UV/vis spectra of 2-azo-thiazole 2 show a satis-
fying E-to Z-conversion of estimated more than 80%, NMR ana-
lysis at higher concentrations (3.52 mM) indicates a consider-
ably lower conversion to the Z-isomer (29%). At even higher
concentrations, (17.6 mM) NMR analysis indicates no photo-
chemical E to Z conversion of 2. Back isomerization with
525 nm shows concentration dependence: NMR analysis indi-
cated just a small Z to E conversion of 10%. Interestingly, con-
centration dependence of heteroazo compounds affected the
half-life rather than the photostationary states (also reported
in literature).44,45 We attribute the concentration dependent
switching behaviour to self-quenching effects and the for-
mation of agglomerates. In line with this notion, examination
of highly diluted solutions (5 µM–100 µM) by UV/vis spec-
troscopy showed no concentration dependency. This concen-
compound 2. For the E-isomer, absorption maxima at λmax
=
Fig. 3 Top: UV spectra of 2-azo-thiazole 2 (50 µM, DMSO, left) and
2-azo-imidazole 3 (10 µM, DMSO, right). The spectra of the E-isomers
are plotted in black and the PSS spectra measured after irradiation with
435/420 nm are plotted in red. The PSS of back isomerization with
525 nm is plotted in blue. Bottom: Absorbance of 2-azo-thiazole 2
(DMSO, 411 nm, left) and 2-azo-imidazole 3 (DMSO, 408 nm right) upon
irradiation with 435/420 nm and 525 nm in an alternating sequence.
Fig. 4 Left: UV spectra of 2-diazocine-thiazole 4 (290 µM, DMSO). The
Z-spectrum is plotted in black and the PSS spectrum measured after
irradiation with 405 nm is plotted in red. Right: Absorbances of 2-diazo-
cine-thiazole 4 (DMSO, 556 nm) after alternating irradiation at 405 nm
and 525 nm in repeated switching cycles.
This journal is © The Royal Society of Chemistry and Owner Societies 2019
Photochem. Photobiol. Sci.

View Article Online
Paper
Photochemical & Photobiological Sciences
Table 2 Photostationary states (PSS) and half-lives (t1/2) of compound
2–4 measured in deuterated DMSO at 27 °C
the mode of binding in the active site on a molecular level, we
aimed at X-ray crystallographic analyses of ligand–protein
complexes.
Compound
Isomer: PSS (λ)
PSS (525 nm)
t1/2
2
3
4
Z: 29% (435 nm)
Z: 85% (420 nm)
E: 47% (405 nm)
E: 81%
E: 88%
Z: 100%
13 mina
2.4 h
3.2 d
X-ray crystallographic analysis of photoswitchable ligand-p38α
MAPK and -CK1δ complexes
p38α. Compound 2 was co-crystallized in complex with p38α
MAPK under ambient lab conditions and the structure was
solved by molecular replacement (Fig. 5). Within the structure,
however, a conformation of the diarylazo moiety was observed
that was neither compatible with the expected angles of the E-
nor the Z-configuration. In fact, the reduced hydrazine 31 was
present, forming an additional H-bond with Gly170. Within
the crystallization solution, dithiothreitol (DTT) was used to
avoid oxidation of the protein. Reducing agents such as DTT
and glutathione (GSH) are known to have the reducing strength
to reduce some azo moieties.46–48 Thus, we hypothesized DTT
to be the responsible agent for the reduction of the diarylazo
moiety. To confirm this experimentally, we used a 3.5 nM solu-
tion of compound 2 in a NMR tube, added 3 mM solution of
DTT in DMSO and investigated the reaction progress by 1H
NMR spectroscopy. After addition of the DTT solution, a second
set of signals immediately emerged providing evidence for the
formation of the hydrazine derivative 31. Furthermore, when a
a Determined using UV/vis spectroscopy.
tration dependent switching behaviour of 2-azo-thiazol 2 has
to be further investigated. However, we conclude that the E to
Z conversion at the very low (µM) concentrations employed in
biological assays, are not hampered by aggregation effects. The
half-life of 13 min of E-2-azo-thiazole 2, however, is probably
too short in terms of common enzymatic assays. In contrast,
2-azo-imidazole 3 has promising photophysical properties with
a E to Z conversion of 85% and a half-life of 2.4 hours. The dia-
zocine functionalized thiazole 4 shows a Z to E conversion of
47% with a rather long half-life of 3.2 days. In contrast to com-
pound 2, compound 3 and 4 showed no concentration depen-
dent photostationary states.
Determination of IC50 values
Having photoswitchable compounds 2–4 as isolated isomers, large excess of DTT was used, the signals of 2 completely van-
or photostationary mixtures, in hand, we set out for enzymatic ished to quantitatively yield the hydrazine derivative 31.
evaluation of the inhibition of p38α MAPK and the structurally Additionally, high resolution mass spectrometry confirmed the
highly related CK1δ. In order to switch towards the photo- formation of the hydrazine derivative 31 and kinetic measure-
stationary states, the DMSO compound stock solutions were ments via UV/vis spectroscopy additionally revealed a very short
irradiated with appropriate wavelengths (Fig. 3 and 4) prior to half-life of 3 min for the reduction (see ESI†).
dilution with water. However, in the actual in vitro kinase
Similar to DTT for p38α assays and crystallization, GSH
assays the situation was not as clear as expected. Compounds (glutathione) as reducing/protein stabilizing agent is deployed
2 and 3 in both the E- and Z-configuration were determined to in the regular setup of the CK1δ assays. In contrast to the
be highly potent inhibitors with IC50 values in the nM range above described NMR experiments with DTT, we were not able
(Table 3) which is in sharp contrast to the design concept and to prove the reduction of compound 2 with GSH. However, by
modelling results. In comparison, 4 showed no significant using a large excess of GSH respective changes could be
inhibition in both configurations up to a concentration of detected in UV/vis spectra. From other studies it is reported
10 µM. Strikingly and in contrast to our E-active/Z-less-active that Z-azo compounds were reduced up to 100-times faster
hypothesis, in the regular p38α MAPK and CK1δ assays under than the corresponding E-azo compounds.49 In line with these
controlled light conditions or in the dark, compound 2 was data, in our study the assays enriched in the Z-isomer contain
even more active when the Z-isomer was enriched via photoi- far more hydrazine 31, which could also provide an expla-
somerization. To elucidate this discrepancy and to characterize nation for the apparent higher inhibitory activity of the
Table 3 Biological activities of inhibitors 2–4. IC50 values were determined (n = 3) for the isolated E-isomers, and for the E/Z-PSS’s at respective
irradiation wavelengths (in parentheses). Abbreviations: n.a. not active up to 10 µM inhibitor concentration, PSS: photostationary state
CK1δ
p38α
Isomer
Irradiation
Compound Wavelength
Maximal inhibition (bottom plateau)
SD/% residual kinase activity
Maximal inhibition (bottom plateau)
SD/% residual kinase activity
IC50 SD/nM
147 38.3 nM
IC50 SD/nM
2
3
4
E (dark)
37.9 0.95
29.1 8.62 nM 4.09 3.45
2.36 0.61 nM −1.75 2.62
83.2 8.68 nM 1.91 5.24
PSS (435 nm) 54.7 9.70 nM 5.40 0.25
E (dark)
PSS (420 nm) 218 2.33 nM
Z (dark) n.a.
PSS (405 nm) n.a.
138 26.5 nM
23.0 0.48
22.4 5.00
—
—
115 28.9 nM
n.a.
n.a.
8.24 2.38
—
—
Photochem. Photobiol. Sci.
This journal is © The Royal Society of Chemistry and Owner Societies 2019

View Article Online
Photochemical & Photobiological Sciences
Paper
Fig. 5 Crystal structure of p38α MAPK in complex with compound 2 and 3. (a) Overview of the kinase domain including the binding pocket and
important regions (hinge region – green, glycine-rich loop – magenta, αC-helix – orange, activation loop – red). (b) Zoom of the binding pocket
highlighting the differences of the mode of binding of compound 2 (and the reduced hydrazine compound 31) and compound 3. The core of the
molecules always binds in a similar orientation, whereas differences can be observed within the diarylazo moiety and the proximal glycine-rich loop.
The solvent exposed 2,5-dimethoxyphenyl-group is not well resolved in the structures and is modelled in different conformations. (c) Most inter-
actions of the compounds 2, 3 and 31 are similar in the different structures and mostly of hydrophobic nature. Observed differences included an
additional interaction of the hydrazine of compound 31 and the carbonyl of Gly170 as well as an additional stacking interaction of the fluorophenyl
group of compound 3 and Tyr35 (glycine-rich loop).
small amount of the molecule was reduced (presumably
because of trace amounts of remaining DTT or the protein
itself being responsible for the reduction).
Table 4 Measured IC50 values of photoswitchable inhibitor 2 with
p38α in the absence of DTT. IC50 values were determined (n = 3) for the
isolated E-isomer, and for the E/Z-PSS after irradiation with 435 nm
In contrast to our observation of reduction of compound 2
during crystallization and within the in vitro kinase assay, com-
pound 3 was stable under conditions containing DTT. The
redox properties of azobenzenes are significantly influenced by
the electronic nature of the substituents. Concerning the
reduction mechanism, electron poor azo compounds are gen-
erally more sensitive to reduction than electron rich
Maximal inhibition (bottom plateau)
Compound 2 IC50 SD/nM SD/% residual kinase activity
Dark (E)
PSS (435 nm) 105 34.4
67.5 23.7
46.6 1.54
9.98 3.71
assumed Z-isomer (in fact being hydrazine 31). This leads to compounds.50–52 This is consistent with our observation, that
the conclusion that the most inhibitory active compound is the 2-azo-thiazole 2 is rather electron poor compared to the
neither the E- nor Z-isomer of 2 but rather the hydrazine com- 2-azo-imidazole 3, thus leading to a reduction of the thiazole
pound 31. This hypothesis was verified by repetition of the compound 2. After successful co-crystallization of compound 3
p38α in vitro kinase assay in the absence of DTT (Table 4). with p38α (Fig. 4), the diarylazo moiety was observed mainly in
Herein, the IC50 values for both the E-isomer and the PSS with the E-configuration. In contrast to the previous example, the
enriched Z-isomer ratio are both higher in absence of DTT, E-diarylazo moiety is still intact in 3, indicating enhanced sus-
and the PSS measurement shows slightly less activity com- ceptibility of thiazole 2 towards reduction, compared to imid-
pared to the isolated E-isomer. However, so far it was not poss- azole 3. Furthermore, 1H NMR and UV/vis investigations of
ible to perform the CK1δ assays without GSH.
imidazole 3 treated with DTT and GSH showed no indication
Based on the evidence for the formation of the hydrazine for a reduction to the respective hydrazine.
31 in presence of DTT, we co-crystallized the complex of com-
CK1δ. We also co-crystallized compound 2 and compound 3
pound 2 and p38α without adding DTT. In this structure we together with CK1δ (Fig. 6). Being aware of the high reduction
observed compound 2 in the E-configuration as the major susceptibility of compound 2, we removed DTT from the CK1δ
population (see ESI Fig. S50c†). However, even in this case a protein sample before setting up crystallization trials with this
This journal is © The Royal Society of Chemistry and Owner Societies 2019
Photochem. Photobiol. Sci.

View Article Online
Paper
Photochemical & Photobiological Sciences
Fig. 6 Crystal structures of CK1δ in complex with compounds 2 and 3. (a) Overview of the CK1δ complex structures with compound 2 or 3 respect-
ively. Important regions are colored: The hinge region in green, the glycine-rich loop in magenta, αC-helix in orange, and the activation loop in red.
(b and d) Close-up of the binding pocket of both complex structures. The electron density (2Fo-Fc map) of the bound compounds is depicted as
blue mesh with a contour level of 1σ. (c and e) 2D representation of most important ligand–protein interactions. The 4,5-diaryl-imidazole scaffold
binds to CK1δ in a similar manner as it was for other compounds observed previously.29 The diarylazo moiety was poorly resolved in the electron
density suggesting that it adopts several conformations when bound to CK1δ.
compound. Herein, the ligand–protein complex structures ent results for inhibitory activities of the respective photo-
revealed that the 4,5-diaryl moieties of both compounds were switchable E/Z-isomers of 2 and 3. Surprisingly, by solving a
binding in an identical way as observed in other inhibitors ligand–protein complex of 2 in p38α by X-ray analysis we dis-
with this scaffold.29 In both data sets, the azo moiety was covered the diarylazo unit being reduced by DTT to a hydra-
present and could be well resolved by the electron density zine moiety yielding the new compound 31. Since reducing
while the phenyl- (compound 2) or fluorophenyl- (compound and stabilizing agents such as DTT and GSH are routinely
3) moieties were barely visible in the electron density maps. used in biological assays our observation might generally
Hence, our data is indicating that the diarylazo moieties of pertain to other photoswitchable diarylazo inhibitors used in
both inhibitors are not adopting a stable conformation in similar studies. Re-examination might be necessary in some
CK1δ. This flexibility could also explain why only minor photo- cases. This is especially true since the higher susceptibility of
induced differences in the inhibition potencies of these com- the Z isomer toward reduction can simulate an apparent
pounds could be observed (Fig. 6).
photoswitchable system and lead to wrong conclusions.
As a second limitation of the photoswitchable activity
approach, we have observed a substantial ligand flexibility of
compound 3 in CK1δ and p38α not translating into sufficient
steric clashes of neither the E- nor the Z-ligand with the
Conclusions
Surprisingly, photoswitchable inhibitors based on diarylazo protein. Consequently, both isomers are able to bind to the
cores often exhibit small differences regarding biological active site, thus not yielding significant differences in inhibi-
activities of their E- and Z-isomers. In this study we aimed at tory activities. In line with this observation and in addition to
the design, synthesis and characterization of photoswitchable the flexibility of the ligand, a conformational adaption of the
kinase inhibitors. As the model system we used the well protein could also be responsible for an induced-fit type of
known vicinal diaryl scaffold (tear-drop binder53,54) of p38α- binding levelling any designed E/Z affinity differences. With
and CK1δ-inhibitors. This basic structure was functionalized these results in mind, future investigations should carefully
with photoswitchable diarylazo- respectively diazocine moi- choose the position of the photoswitch within the core of the
eties in such a way that the corresponding E- and Z-isomers pharmacophore rather than being placed in a (flexible) side
were predicted to exhibit different inhibitory activities. Much chain. Furthermore, compound concentration, PSS and
to our surprise, the diazocine-based compound 4 turned out respective E/Z half-lives in an aqueous medium play crucial
not to inhibit these kinases, whereas compounds 2 and 3 were roles for effective photoswitches in biological assays. Finally,
potent inhibitors with IC50 values in the nanomolar range. the assay set-up under controlled light conditions turned out
However, in the course of our studies, we observed inconsist- to be highly important to avoid artefacts.
Photochem. Photobiol. Sci.
This journal is © The Royal Society of Chemistry and Owner Societies 2019

View Article Online
Photochemical & Photobiological Sciences
Paper
Consideration of these pitfalls and limitations in photo-
switchable inhibitor design might improve efficiencies and
reduce the probability of misinterpretations.
2 W. Browne and B. L. Feringa, Nat. Nanotechnol., 2006, 1,
25–35.
3 S. Kassem, T. van Leeuwen, A. S. Lubbe, M. R. Wilson,
B. L. Feringa and D. A. Leigh, Chem. Soc. Rev., 2017, 46,
2592–2621.
4 K. Matsuda and M. Irie, J. Photochem. Photobiol., C, 2004, 5,
169–182.
Notes
5 T.-T.-T. Nguyen, D. Türp, M. Wagner and K. Müllen, Angew.
Chem., Int. Ed., 2013, 52, 669–673.
6 G. Mayer and A. Heckel, Angew. Chem., Int. Ed., 2006, 45,
4900–4921.
7 A. A. Beharry and G. A. Woolley, Chem. Soc. Rev., 2011, 40,
4422–4437.
Compound
PDB code p38α
PDB code CK1δ
2
3
31
6HWU
6HWV
6HWT
6HMR
6HMP
—
8 M. M. Lerch, M. J. Hansen, G. M. van Dam, W. Szymanski
and B. L. Feringa, Angew. Chem., Int. Ed., 2016, 55,
10978.
9 K. Hüll, J. Morstein and D. Trauner, Chem. Rev., 2018, 118,
10710–10747.
10 W. Szymanski, M. E. Ourailidou, W. A. Velema, F. J. Dekker
and B. L. Feringa, Chem. – Eur. J., 2015, 21, 16517–16524.
11 D. Vomasta, C. Högner, N. R. Branda and B. König, Angew.
Chem., Int. Ed., 2008, 47, 7644–7647.
Authors will release the atomic coordinates and experi-
mental data upon article publication. Co-crystallization of
compounds 2, 3, and 31 in p38α; compounds 2 and 3 in CK1δ.
CCDC 1885848† contains the supplementary crystallo-
graphic data for compound 13.
Conflicts of interest
12 R. Ferreira, J. R. Nilsson, C. Solano, J. Andrésson and
M. Grøtli, Sci. Rep., 2015, 5, 1–8.
There are no conflicts to declare.
13 J. Kuil, L. T. M. van Wandelen, N. J. de Mol and
R. M. J. Liskamp, Bioorg. Med. Chem., 2008, 16, 1393–1399.
14 A. D. Abell, M. A. Jones, A. T. Neffe, S. G. Aitken, T. P. Cain,
R. J. Payne, S. B. McNabb, J. M. Coxon, B. G. Stuart,
D. Pearson, H. Y.-Y. Lee and J. D. Morton, J. Med. Chem.,
2007, 50, 2916–2920.
15 C. Falencyk, M. Schiedel, B. Karaman, T. Rumpf,
N. Kuzmanovic, M. Grøtli, W. Sippl, M. Jung and B. König,
Chem. Sci., 2014, 5, 4794.
16 M. J. Hansen, W. A. Velema, G. de Bruin, H. S. Overkleeft,
W. Szymanski and B. L. Feringa, ChemBioChem, 2014, 15,
2053–2057.
17 C. E. Weston, A. Krämer, F. Colin, Ö. Yildiz, M. G. J. Baud,
F.-J. Meyer-Almes and M. J. Fuchter, ACS Infect. Dis., 2017,
3, 152–161.
18 D. Wutz, D. Gluhacevic, A. Chakrabarti, K. Schmidtkunz,
D. Robaa, F. Erdmann, C. Romier, W. Sippl, M. Jung and
B. König, Org. Biomol. Chem., 2017, 15, 4882–4896.
19 N. A. Smith, L.-M. Altmann, N. Wössner, E. Bauer, M. Jung
and B. König, J. Org. Chem., 2018, 83, 7919–7927.
20 D. Wilson, J. Li and N. Branda, ChemMedChem, 2017, 12,
284–287.
21 C. Schwab, A. J. DeMaggio, N. Ghoshal, L. I. Binder,
J. Kuret and P. L. McGeer, Neurobiol. Aging, 2000, 21, 503–
510.
22 G. Manning, D. B. Whyte, R. Martinez, T. Hunter and
S. Sudarsanam, Science, 2002, 298, 1912–1934.
23 J. Zhang, P. L. Yang and N. S. Gray, Nature, 2009, 9, 28–39.
24 A. C. Dar and K. M. Shokat, Annu. Rev. Biochem., 2011, 80,
769–795.
Acknowledgements
The authors from Kiel university acknowledge financial support
by the Deutsche Forschungsgesellschaft (DFG) within the
Sonderforschungsbereich 677, “Function by Switching”. Chiara
Ianes is financially supported by a grant of the medical faculty of
Ulm University (“Bausteinprogramm”) awarded to Joachim
Bischof. Work in the lab of Uwe Knippschild was supported by
the DFG (KN356/6-1, and SFB 1149, project B04). Daniel Rauh is
thankful for support from the German Federal Ministry for
Education and Research (NGFNPlus and e:Med) (Grant No. BMBF
01GS08104, 01ZX1303C), the Deutsche Forschungsgemeinschaft
(DFG) and the German federal state North Rhine Westphalia
(NRW) and the European Union (European Regional
Development Fund: Investing In Your Future) (EFRE-800400). The
crystallographic experiments leading to these results have received
funding from the European Community’s Seventh Framework
Programme (FP7/2007-2013) under grant agreement No. 283570
(BioStruct-X). CK1δ crystals were grown in the Cologne
Crystallization facility (http://C2f.uni-koeln.de/). We thank the staff
at the Swiss Light Source, Paul Scherrer Institute, Villigen,
Switzerland for their support during data collection. We gratefully
acknowledge the help of Dr Dieter Schollmeyer, University of
Mainz, Institute for Organic Chemistry, Germany, for X-ray ana-
lysis of compound 13.
Notes and references
1 V. Balzani, A. Credi, F. M. Raymo and J. F. Stoddart, Angew. 25 S. Gross, R. Rahal, N. Stransky, C. Lengauer and
Chem., Int. Ed., 2000, 39, 3348–3391.
K. P. Hoeflich, J. Clin. Invest., 2015, 125, 1780–1789.
This journal is © The Royal Society of Chemistry and Owner Societies 2019
Photochem. Photobiol. Sci.

View Article Online
Paper
Photochemical & Photobiological Sciences
26 M. Huang, A. Shen, J. Ding and M. Geng, Trends Pharmacol. 40 J. Roger and H. Doucet, Tetrahedron, 2009, 65, 9772–9781.
Sci., 2014, 35, 41–50.
27 C. Sun and R. Bernards, Trends Biochem. Sci., 2014, 39,
465–474.
41 M. Hammerich, C. Schütt, C. Stähler, P. Lentes,
F. Röhricht, R. Höppner and R. Herges, J. Am. Chem. Soc.,
2016, 138, 13111–13114.
28 S. Klaeger, S. Heinzlmeir, M. Willhelm, et al., Science, 2017, 42 H. Kim, I. Yang, R. S. Patil, S. Kang, J. Lee, H. Choi,
358, 1–16.
M.-S. Kim, S.-J. Nam and H. Kang, J. Nat. Prod., 2014, 77,
2716–2719.
29 J. Halekotte, L. Witt, C. Ianes, M. Krüger, M. Bührmann,
D. Rauh, C. Pichlo, E. Brunstein, A. Luxenburger, 43 D. K. Joshi, M. J. Mitchell, D. Bruce, A. J. Lough and
U. Baumann, U. Knippschild, J. Bischof and C. Peifer, mole-
cules, 2017, 22, 522.
30 C. E. Weston, R. D. Richardson, P. R. Haycock,
A. J. P. White and M. J. Fuchter, J. Am. Chem. Soc., 2014,
136, 11878–11881.
H. Yan, Tetrahedron, 2012, 68, 8670–8676.
44 J. Calbo, C. E. Weston, A. J. P. White, H. S. Rzepa,
J. Contreras-García and M. J. Fuchter, J. Am. Chem. Soc.,
2017, 139, 1261–1274.
45 S. Devi, M. Saraswat, S. Grewal and S. Venkataramani,
J. Org. Chem., 2018, 83, 4307–4322.
31 J. Otsuki, K. Suwa, K. K. Sarker and J. Sinha, J. Phys. Chem.
A, 2007, 111, 1403–1409.
46 W. G. Levine, Drug Metab. Rev., 1991, 23, 253–309.
32 R. Siewertsen, H. Neumann, B. Buchheim-Stehn, 47 A. A. Beharry, L. Wong, V. Tropepe and G. A. Woolley,
R. Herges, C. Näther, F. Renth and F. Temps, J. Am. Chem.
Soc., 2009, 131, 15594–15595.
33 S. Laufer and P. Koch, Org. Biomol. Chem., 2008, 6, 437–439.
Angew. Chem., Int. Ed., 2011, 50, 1325–1327.
48 L. Stricker, M. Böckmann, T. M. Kirse, N. L. Doltsinis and
B. J. Ravoo, Chem. – Eur. J., 2018, 24, 8639.
34 S. Laufer and A. J. Liedtke, Tetrahedron Lett., 2006, 47, 49 C. Renner and L. Moroder, ChemBioChem, 2006, 7, 868–
7199–7203.
878.
35 H. Beyer and G. Henseke, Chem. Ber., 1950, 88, 1233–1236.
50 S. Zbaida, Drug Metab. Rev., 1995, 27, 497–516.
36 B. Sahlmann, PhD Thesis, Christian-Albrechts-University 51 C. Boulègue, M. Löweneck, C. Renner and L. Moroder,
Kiel, 2013. ChemBioChem, 2007, 8, 591–594.
37 M. Dommaschk, M. Peters, F. Gutzeit, C. Schütt, C. Näther, 52 W. A. Velema, W. Szymanski and B. L. Feringa, J. Am.
F. D. Sönnichsen, S. Tiwari, C. Riedel, S. Boretius and
R. Herges, J. Am. Chem. Soc., 2015, 137, 7552–7555.
38 M. J. Hansen, M. M. Lerch, W. Szymanski and
B. L. Feringa, Angew. Chem., Int. Ed., 2016, 55, 13514–
13518.
Chem. Soc., 2014, 136, 2178–2191.
53 B. A. Thaher, M. Arnsmann, F. Totzke, J. E. Ehlert,
M. H. G. Kubbutat, C. Schächtele, M. O. Zimmermann,
P. Koch, F. M. Boeckler and S. A. Laufer, J. Med. Chem.,
2012, 55, 961–965.
39 T. Wendler, C. Schütt, C. Näther and R. Herges, J. Org. 54 A. Freitag and S. A. Laufer, Nachr. Chem., 2015, 63, 420–
Chem., 2012, 77, 3284–3287.
425.
Photochem. Photobiol. Sci.
This journal is © The Royal Society of Chemistry and Owner Societies 2019