Photoswitchable azo‐ and diazocine‐functionalized derivatives of the vegfr‐2 inhibitor axitinib

Article abstract of DOI:10.3390/ijms21238961

In this study, we aimed at the application of the concept of photopharmacology to the approved vascular endothelial growth factor receptor (VEGFR)‐2 kinase inhibitor axitinib. In a previous study, we found out that the photoisomerization of axitinib’s stilbene‐like double bond is unidirectional in aqueous solution due to a competing irreversible [2+2]‐cycloaddition. Therefore, we next set out to azologize axitinib by means of incorporating azobenzenes as well as diazocine moieties as photoresponsive elements. Conceptually, diazocines (bridged azobenzenes) show favorable photoswitching properties compared to standard azobenzenes because the thermodynamically stable Z‐isomer usually is bioinactive, and back isomerization from the bioactive E‐isomer occurs thermally. Here, we report on the development of different sulfur– diazocines and carbon–diazocines attached to the axitinib pharmacophore that allow switching the VEGFR‐2 activity reversibly. For the best sulfur–diazocine, we could verify in a VEGFR‐2 kinase assay that the Z‐isomer is biologically inactive (IC50 >> 10,000 nM), while significant VEGFR‐2 inhibition can be observed after irradiation with blue light (405 nm), resulting in an IC50 value of 214 nM. In summary, we could successfully develop reversibly photoswitchable kinase inhibitors that exhibit more than 40‐fold differences in biological activities upon irradiation. Moreover, we demonstrate the potential advantage of diazocine photoswitches over standard azobenzenes.

Full text of DOI:10.3390/ijms21238961

International Journal of
Molecular Sciences
Article
Photoswitchable Azo- and Diazocine-Functionalized
Derivatives of the VEGFR-2 Inhibitor Axitinib
Linda Heintze 1 , Dorian Schmidt , Theo Rodat 1 , Lydia Witt , Julia Ewert , Malte Kriegs ,
,†
1,†
1
2
3
Rainer Herges 2 and Christian Peifer *
1,
1
Institute of Pharmacy, Christian-Albrechts-University of Kiel, 24118 Kiel, Germany;
lheintze@pharmazie.uni-kiel.de (L.H.); dschmidt@pharmazie.uni-kiel.de (D.S.);
trodat@pharmazie.uni-kiel.de (T.R.); lwitt@pharmazie.uni-kiel.de (L.W.)
Otto-Diels-Institute of Organic Chemistry, Christian-Albrechts-University of Kiel, 24098 Kiel, Germany;
2
jewert@oc.uni-kiel.de (J.E.); rherges@oc.uni-kiel.de (R.H.)
Laboratory of Radiobiology & Experimental Radiooncology and UCCH Kinomics Core Facility,
3
University Medical Center Hamburg-Eppendorf, 20246 Hamburg, Germany; m.kriegs@uke.de
*
Correspondence: cpeifer@pharmazie.uni-kiel.de; Tel.: +49-431-880-1137
†
These authors contributed equally to this work.
ꢀ
ꢁꢂꢀꢃꢄꢅꢆꢇ
Received: 28 September 2020; Accepted: 22 November 2020; Published: 25 November 2020
ꢀ
ꢁꢂꢃꢄꢅꢆ
Abstract: In this study, we aimed at the application of the concept of photopharmacology to
the approved vascular endothelial growth factor receptor (VEGFR)-2 kinase inhibitor axitinib.
In a previous study, we found out that the photoisomerization of axitinib’s stilbene-like double
bond is unidirectional in aqueous solution due to a competing irreversible [2+2]-cycloaddition.
Therefore, we next set out to azologize axitinib by means of incorporating azobenzenes as well as
diazocine moieties as photoresponsive elements. Conceptually, diazocines (bridged azobenzenes)
show favorable photoswitching properties compared to standard azobenzenes because the
thermodynamically stable Z-isomer usually is bioinactive, and back isomerization from the bioactive
E-isomer occurs thermally. Here, we report on the development of different sulfur–diazocines
and carbon–diazocines attached to the axitinib pharmacophore that allow switching the VEGFR-2
activity reversibly. For the best sulfur–diazocine, we could verify in a VEGFR-2 kinase assay that the
Z-isomer is biologically inactive (IC50 >> 10,000 nM), while significant VEGFR-2 inhibition can be
observed after irradiation with blue light (405 nm), resulting in an IC50 value of 214 nM. In summary,
we could successfully develop reversibly photoswitchable kinase inhibitors that exhibit more than
40-fold differences in biological activities upon irradiation. Moreover, we demonstrate the potential
advantage of diazocine photoswitches over standard azobenzenes.
Keywords: photopharmacology; diazocine; azobenzene; photoswitchable kinase inhibitor;
axitinib; VEGFR
1
. Introduction
Photopharmacology aims at the spatial and temporal control of the pharmacological activity of a
photoresponsive compound in biological settings by irradiation with light. Thereby, the disadvantages
of classical pharmacotherapy such as severe side effects, emergence of resistances, and environmental
contamination could be reduced [1,2]. For such a photopharmacological approach, photoresponsive
groups have to be incorporated into a bioactive molecule. For example, light-induced E/Z isomerizations
or ring-closing reactions can be used to switch the configurational/constitutional states of a drug.
For reversible photoswitching, irradiation at two different wavelengths is necessary to allow a selective
excitation of each state. For molecules with a metastable state, back-switching to the thermodynamically
Int. J. Mol. Sci. 2020, 21, 8961; doi:10.3390/ijms21238961
www.mdpi.com/journal/ijms

Int. J. Mol. Sci. 2020, 21, 8961
2 of 20
stable form can also occur thermally. From a pharmacological perspective, one state should ideally be
bioinactive, while the other state should be significantly more active. In addition, a high photoconversion
from the inactive state to the biologically active state (and vice versa) is mandatory to achieve robust
effects in biological assays. For basic in vitro applications, it is desirable to obtain at least a factor of 10,
but rather a factor of 100, regarding the biological activity in the thermodynamic equilibrium and the
photostationary state (PSS), respectively [3–5].
In addition to these stringent requirements, for in vivo applications, lead-like or even drug-like
physicochemical parameters of a compound need to be met, which are typically optimized in medicinal
chemistry projects. Against this background, we develop reversibly photoswitchable compounds
on the basis of an established bioactive pharmacophore. Recently, our group reported on the
photoswitching of the kinase inhibitor axitinib. Axitinib is inherently photoswitchable because it
includes a photoisomerizable C=C double bond. Due to their essential role in cellular signaling,
kinases have become interesting drug targets. So far, 55 small molecule kinase inhibitors have been
approved by the FDA (Food and Drug Administration USA) [
inhibitors may entail problems such as severe side effects or the development of resistances, which could
be reduced by a more focused approach involving photoresponsive kinase inhibitors [ ]. However,
6,7]. An oral therapy involving kinase
8
previously reported photoswitchable small molecule kinase inhibitors (Figure 1) exhibited only slight
differences in the biological activity between the metastable, irradiated, and the thermodynamically
stable state [9]. These examples include three azobenzene-based kinase inhibitors with a 1.6- to 11-fold
difference in biological activities upon photoswitching [10
–12] and a protein kinase C inhibitor based
on a diarylethene-like maleimide showing a higher difference with a factor of 26, but no reversible
switching in aqueous solutions (similar to axitinib (1), see next paragraph) [13].
Figure 1. Previouslyreportedphotoswitchablesmallmoleculekinaseinhibitors. Published azobenzene-based
kinase inhibitors exhibited only slight differences (1.6- to 11-fold) in the biological activity before and
after irradiation [10–12]. Larger differences but no reversible photoswitching in aqueous solutions
was shown for a maleimide-based PKC (Protein Kinase C) inhibitor [13] as well as for the approved
VEGFR-2 inhibitor axitinib [14].

Int. J. Mol. Sci. 2020, 21, 8961
3 of 20
Axitinib as Template for Photoresponsive Kinase Inhibitors
Axitinib (
It targets receptor tyrosine kinases, especially vascular endothelial growth factor receptor (VEGFR)
–3, which play important roles e.g., in angiogenesis and tumor growth [15]. In our previous study,
1, Figure 1) has been licensed for second-line therapy of renal cell carcinoma since 2012.
1
we were able to photoswitch axitinib, which contains a stilbene-like double bond allowing for E/Z
isomerization. For Z-axitinib and E-axitinib, a 43-fold difference in their respective biological activities
could be demonstrated. Unfortunately, the photoswitching of axitinib (1) is not reversible in aqueous
solution due to a competing [2+2]-cycloaddition yielding a bioinactive dimer. Hence, we concluded
that the original stilbene-like system of axitinib (1) is unsuitable as a reversible photoswitch, at least in
aqueous solution [14].
Therefore, in this study, we developed derivatives of axitinib (
features. The most obvious concept to avoid a [2+2]-photocycloaddition of axitinib (
the C=C double bond by an azo group. This approach has been successfully applied to several
other compounds before and was named “azologization” by the Trauner group [ 16]. Indeed,
1) showing optimized photochemical
1) is to replace
3,
some azo-derivatives of axitinib have already been reported by Wei et al., about two years ago; however,
there are no specific data on the photoswitching properties of these compounds [17].
We have also synthesized “azoaxitinib” (
assumed that E- should be biologically active analogous to E-axitinib (
Z-azoaxitinib (Z- ) to the VEGF receptor should be sterically hindered. However, after synthesis,
we could not observe the photoswitching of E-azoaxitinib (E- ) by classical UV/VIS spectroscopy
Figure S2) and concluded that the Z-isomer Z-2a is in equilibrium with hydrazone 2b, resulting in
very fast thermal back isomerization (see Figure 2). This phenomenon has been reported for similar
indole and imidazole compounds [20 22]. Interestingly, we observed a quite different behavior of
2
, Scheme S1 [18
,
19] and Figure S9) previously and
2
1
), whereas the binding of
2
2
(
–
isomerization of compound 12 during the synthesis of “azoaxitinib” (Scheme S1). Compound 12
appears to be stabilized as a 2H/Z tautomer by an intramolecular H-bond. A similar tautomerism
within the heterocycle was found for 3(5)-arylazopyrazoles [23]. However, this could not be observed
for the final product
E-2a/E-2b) in a VEGFR-2 kinase assay revealed a decrease of the inhibitory activity (IC50 = 415 nM,
Figure S8) compared to E-axitinib (IC50 = 19 nM) [14]. In fact, a simple azologization of axitinib is
basically only partly useful, as azoaxitinib ( ) is considered bioactive in the thermodynamically stable
2. Moreover, biological evaluation of the tautomer mixture of E-azoaxitinib
(
2
E-form. Hence, a photoresponsive approach would only allow a temporal reduction of the bioactivity
of E-azoaxitinib to some degree.
From our point of view, a more expedient way to develop photoswitchable axitinib derivatives is
the implementation of diazocines (bridged azobenzenes), which are thermodynamically stable in the
(
bioinactive) Z-form and show higher photoconversion rates compared to classical azobenzenes [24].
Up to now, there are only a few but successful examples of photoswitchable drugs containing
diazocines as the switching unit. Diazocines were used to photoswitch peptide conformations [25],
DNA hybridization [26], and potassium channel activities [27,28]. Based on these examples, we designed
and synthesized diazocine-functionalized derivatives of axitinib. Conceptually, these photoresponsive
compounds were designed to be inactive in the thermodynamically stable diazocine Z-configuration
but bioactive in the metastable and photoinduced E-configuration.

Int. J. Mol. Sci. 2020, 21, 8961
4 of 20
Figure 2. Azo-hydrazone tautomerization of azoaxitinib (2). Due to the tautomeric equilibrium, the azo
double bond is temporally turned into a free rotating single bond. This allows for fast relaxation
from Z-isomer (Z- a) to the thermodynamically stable E-isomer (E- a) by rotation and subsequently
back tautomerization.
2
2
2
. Results
In order to develop reversibly photoswitchable kinase inhibitors derived from axitinib (
we designed, synthesized, and characterized three different diazocine-functionalized derivatives (5–7)
1),
and compared them with the corresponding classical azobenzene derivatives (
an overview of the compounds and related results discussed below.
3 and 4). Table 1 gives
2
.1. Molecular Modeling
2.1.1. Molecular Modeling of Azobenzene Derivatives 3 and 4 in the ATP Binding Pocket of VEGFR-2
Instead of connecting the azo group directly to the indazole scaffold of axitinib as in compound 2,
we appended an azobenzene moiety to the pharmacophore in order to prevent the above described
azo-hydrazone tautomerization. Thereby, the azobenzene moiety was either linked in the meta- or
para-position of the azo-group (compounds
studies were performed for both azobenzene derivatives
pdb: 4AG8) [29]. For each configuration, that is to say for the E-isomers as well as for the Z-isomers,
reasonable binding modes were found. Figure 3 shows the binding modes of meta-azobenzene-axitinib
), which is representative for both azobenzene derivatives. Calculated Glide docking scores [30 32
as an assessment of the binding affinities are also similar for both isomers respectively (E- 13.9, Z-
14.0; E- 13.7, Z- 13.8). These modeling results suggested that photoswitching will probably not
significantly affect the biological activities of both isomers.
3
and
4
, Table 1). Prior to synthesis, molecular modeling
3
and within the binding pocket of VEGFR-2
4
(
(3
–
]
3:
−
3:
−
4:
−
4:
−

Int. J. Mol. Sci. 2020, 21, 8961
5 of 20
Table 1. Summary of the synthesized axitinib derivatives and their photochemical and biological
characterization. λexc: irradiation wavelength, PSS: photostationary state, t1/2: half-life of the
metastable form.
◦
t1/2 (h), 37 C,
PSS (E/Z%),
DMSO (NMR)
#
1
Residue (R)
λexc (nm)
DMSO
IC50 VEGFR-2 (nM)
(
UV/VIS)
E: 19
Z: 817
PSS 385 nm: 29
3
85 (Z→E)
88/12 (Z→E)
51/49 (E→Z)
>>12 *
3
65 (E→Z)
2
3
n/a
n/a
n/a
E: 415
E: 1077 **
PSS 365 nm: 1289
365
17/83
43.1
E: 1020 **
PSS 385 nm: 1435
4
5
385
405
20/80
47/53
5.7
7.3
Z: >10,000
PSS 405 nm: 214
Z: >10,000
PSS 405 nm: 251
6
7
405
405
25/75
60/40
3.7
1.5
Z: n/a **
PSS 405 nm: 493
*
No thermal back isomerization could be observed over 12 h. ** residual kinase activity ≥ 33%.
a
b
Figure 3. Calculated binding mode of meta-linked azobenzene-axitinib (3) in vascular endothelial
growth factor receptor (VEGFR)-2 (pdb: 4AG8) [29]. Protein surface displayed in gray. For the
azobenzene, a plausible binding mode is found in E-configuration (
well as in Z-configuration (b), Glide docking score: −14.0).
a), Glide docking score:
−13.9) as

Int. J. Mol. Sci. 2020, 21, 8961
6 of 20
2
.1.2. Molecular Modeling of Diazocine Derivatives 5–7 in the ATP Binding Pocket of VEGFR-2
In the light of the above-mentioned modeling results, we further attached a photoresponsive
element to the pharmacophore of axitinib that causes a more distinct structural change of the molecule
geometry upon irradiation. In this context, diazocines are well suited because the thermodynamically
stable Z-isomer is considered to be rather biologically inactive due to steric hindrance of the angled
Z-structure at the receptor binding pocket (Figure 4). Upon isomerization toward the more leveled
E-isomer, a major change in the molecule’s total structure is caused. The resulting more planar
E-isomer should be able to bind to the target protein. Furthermore, diazocines were described
to have suitable switching properties with high switching efficiencies and red shifted excitation
wavelengths compared to the azobenzenes [24
sulfur-diazocine-functionalized axitinib derivatives linked in meta (
Furthermore, we investigated an ethylene-bridged diazocine ( ) as a carbon-analogue compound,
which usually show slightly higher conversion rates compared to the sulfur–diazocines.
,33]. For these reasons, we designed and modeled two
5
) or para ( ) of the azo-group.
6
7
a
b
Figure 4.
(a) Calculated binding mode of S-diazocine-functionalized axitinib (5) in E-configuration
(
chair) to VEGFR-2 (pdb: 4AG8) [29]. Shown is the meta-linked derivative representative for both
sulfur–diazocine compounds. ( ) Superposition of Z- and VEGFR-2 binding pocket. While retaining the
b
5
hydrogen bonds of the pharmacophore, the Z-diazocine moiety clashes with the protein. Protein surface
displayed in gray. Yellow dotted lines: hydrogen bonds; light blue dotted lines: π–π interactions.
In line with the above-mentioned concept, docking analyses of the diazocine compounds in the
ATP binding pocket of VEGFR-2 (pdb: 4AG8) [29] revealed plausible binding modes for the stretched
E-isomers of all three derivatives ( ). As expected, no compelling binding modes for the Z-isomers
could be found. Figure 4 displays the pose of the meta-substituted sulfur–diazocine ( ) in the binding
5–7
5
pocket of VEGFR-2, which is representative for both sulfur–diazocine derivatives. Figure 4a shows the
calculated binding mode of the E-isomer, 4b illustrates the steric clash of the Z-isomer in a superposition
with the protein (the corresponding target interactions of carbon–diazocine (7) to key residues of the
binding site are shown in Figure S1).
In addition to the docking experiments in which the receptor binding pocket grid is considered to
be rigid, we also performed induced-fit docking calculations [34] for compounds . In the induced-fit
5–7
model, all amino acid residues in a distance of 5 Å to the ligand are set to be conformationally flexible.
Table 2 gives an overview of the docking scores obtained for both rigid Glide docking and induced-fit
docking. Interestingly, in the induced-fit mode, a plausible binding pose is also found for the Z-isomer
of the carbon–diazocine (7), but not for the sulfur–diazocines 5 and 6. This finding could be explained
by the fact that the angle between both phenyl rings of the Z-diazocine configuration is significantly
◦
◦
larger for the carbon–diazocine (85.8 ) compared to the sulfur–diazocine (76.1 ). Actually, the tighter

Int. J. Mol. Sci. 2020, 21, 8961
7 of 20
sulfur–diazocines clash even with the flexible protein backbone, while the binding pocket residues are
able to adapt to the larger carbon–diazocine ligand (Figure 5).
Table 2. Overview of the docking results of diazocine-functionalized axitinib derivatives. Calculated
docking scores from Glide and induced-fit docking are given for E- and Z-isomer, respectively. As a
necessary condition for an allowed binding mode, all four hydrogen bonds of the axitinib pharmacophore
were defined (H-bond constraints). For the induced-fit docking, all amino acids in a distance of 5 Å
were considered flexible.
Glide Score
Induced-Fit Score
#
5
Residue (R)
E-Isomer a
Z-Isomer
E-Isomer a
Z-Isomer
−14.1
ꢀ
−16.2
ꢀ
6
7
−14.0
−14.3
ꢀ
ꢀ
−15.5
−16.4
ꢀ
−16.0
a
chair conformation; ꢀ: No plausible binding mode was found.
a
b
Figure 5. (a) B3LYP/631-G*-optimized structures of sulfur- and carbon-bridged diazocines. The planes
◦
◦
opened up by the phenyl rings build an angle of 76.1 and 85.8 , respectively. (
mode of carbon-bridged diazocine Z- (blue) in the binding pocket of VEGFR-2 (PDB: 4AG8) [29
superimposed with the corresponding sulfur–diazocine Z- (green). The superposition of
sulfur–diazocine Z- with the adapted protein of the induced-fit calculation results in a steric clash.
The protein surface is displayed in gray.
b) Induced-fit docking
7
]
5
5
2
.2. Synthesis
For the synthesis of the diazocine-functionalized axitinib derivatives
development of a convergent synthetic route due to the typically low reaction yields of diazocines [24
5
–7
, we aimed at the
33].
,

Int. J. Mol. Sci. 2020, 21, 8961
8 of 20
For this purpose, we used acetyl-protected indazole 24 as common building block and tried to
cross-couple metalated diazocines and azobenzenes in the final reaction step. Building block 24 is also
used in the industrial synthesis of axitinib (1) and has been successfully deployed as an electrophile
in cross-couplings [35]. The synthesis of building block 24 was performed according to published
procedures using methyl 2-sulfanylbenzoate (14) and 6-aminoindazole (45) as starting materials
(Scheme S2) [35–37].
The metalation of azo compounds is generally challenging due to their susceptibility to reduction,
e.g., the use of lithium, zinc, or magnesium bases often leads to the formation of the corresponding
hydrazine compounds [38 39]. For this reason, we used a stannylation protocol reported to be suitable for
azobenzenes [40]. As a start, we tested this procedure for the synthesis of compounds and (Scheme 1).
,
3
4
First, 3-iodoaniline (18) or 4-iodoaniline (19) respectively were converted with nitrosobenzene in a
Baeyer–Mills reaction to form iodinated azobenzenes 20 and 21 [41]. The stannylation was finally
realized in a microwave reaction using hexamethylditin and Pd(PPh ) in dry toluene. The Stille
3
4
◦
reaction of building block 24 and stannylated azobenzenes 22/23 finally succeeded in DMF at 120 C
and the use of Pd(PPh ) as catalyst [40]. Under these conditions, the acetyl protecting group is
3
4
cleaved simultaneously.
Scheme 1. Synthesis of azobenzene-functionalized axitinib derivatives
3
and
4
. Reaction conditions:
®
(
(
i
) aq. KHSO solution (Oxone ), DCM, RT, 6 h; (ii) glacial acetic acid, RT, 24 h, meta: 37%, para: 64%;
3
◦
iii) Sn Me , Pd(PPh ) , dry toluene, 100 C, microwave, 3 h, meta: 86%, para: 62%; (iv) Pd(PPh3)4,
2
6
3 4
◦
DMF, 120 C, 24 h, meta: 52%, para: 32%. (meta/para position of the azo group).
With the standard azobenzene derivatives
and Stille reaction for the synthesis of the diazocine-functionalized derivatives
3
and
4
in hand, we tried to adopt the stannylation
. The required
5–7
halogenated diazocines 35 and 43 were synthesized using two different synthetic routes (Scheme 2 vs.
Scheme 3). The sulfur–diazocines 35 and 36 were synthesized according to the procedure reported
by Schehr using an intramolecular Baeyer–Mills reaction [42]. For the metalation of the diazocines,
we slightly modified the stannylation protocol described above. Compared to the original procedure
◦
of Strueben et al., the temperature was reduced from 150 to 100 C, and the reaction time was extended
from 15 min to 4 h. The stannylated diazocines are relatively stable and were obtained in moderate
yields after purification on silica gel.

Int. J. Mol. Sci. 2020, 21, 8961
9 of 20
Br
CH3
NO2
CH3
NO2
NO2
i)
ii)
H N
2
I
I
2
2
6 meta
7 para
28 meta
29 para
25
H N
2
S
S
iii)
NH2
S
S
S
iv)
30
v)
NH2
NH2
I
I
I
N N
N
O
O N
2
3
3
5 meta
6 para
33 meta
34 para
31 meta
32 para
H
N
O
H C
H
3
vi)
H
N
O
O
S
H C
3
CH3
N
N
S
N
N
S
vii)
N
N
24
I
SnMe3
N N
S
5
6
meta
para
3
3
7 meta
8 para
Scheme 2. Synthesis of metallated sulfur–diazocines 37
/
38 and subsequent Stille coupling. Reaction
◦
conditions: (
i) 1. NaNO , H SO , water, 0 C, 30 min, 2. KI, RT, 16 h; (ii) NBS, DBPO, CCl , reflux,
2
2
4
4
0
6–63 h, meta: 30%, para: 19%; (iii) 1. 2,2 -disulfanediyldianiline (30), NaBH , THF, reflux, 1.5 h, 2.
4
1
2
◦ ◦
8/29, 40 C, 3 h, meta: 56%, para: 67%; (iv) 1. Zinc, NH Cl, water/ethanol, 60 C, 30 min, 2. FeCl ,
4 3
◦
water/ethanol, 0 C RT, 30 min; (v) glacial acetic acid, RT, 16 h, meta: 18%, para: 26%; (vi) Sn Me ,
2 6
◦
◦
Pd(PPh ) , dry toluene, 100 C, microwave, 4 h, meta: 71%, para: 60%; (vii) Pd(PPh ) , DMF, 120 C,
3
4
3 4
2
4 h, meta: 16%, para: 17%. (meta/para position of nitro or azo group respectively).
The synthesis of the mono-iodinated carbon-diazocine 43 turned out to be more complicated
compared to the corresponding sulfur–diazocines 35 and 36 since the ethylene bridge is formed
radically [43,44]. The radical reaction mechanism leads to a mixture of unsubstituted product 41,
mono-substituted product 40 and di-substituted product (not shown), which are extremely difficult to
separate. Hence, the product mixture was used in the next step without purification. The reduction to
the corresponding amino compounds (i.e., 42) succeeded with tin(II) chloride in refluxing ethyl acetate,
whereas a standard hydrogenation using palladium catalyst resulted in dehalogenation. For the ring
closure to the diazocine 43, we used meta-chloroperoxybenzoic acid in acetic acid as reported by the
Trauner group recently [45]. The stannylation of diazocine 44 and subsequent cross-coupling with
building block 24 proceeded as described above.
2
.3. Photochemical Characterization of the Photoswitchable Compounds 3–7
In order to determine the isomer ratios of the PSS as well as to identify the optimal excitation
wavelengths for the E/Z isomerizations, compounds 3–7 were characterized by UV/VIS and NMR
spectroscopy. Due to their poor water solubility along with the low molar absorption coefficients of
diazocines, all compounds were initially analyzed in DMSO.

Int. J. Mol. Sci. 2020, 21, 8961
10 of 20
Scheme 3.
Synthesis of metallated carbon–diazocine 44 and subsequent Stille coupling.
◦ ◦
Reaction conditions: (i) 1. KOtBu, THF, 0 C, 2 min, 2. Br , 0 C, 5 min; (ii) SnCl , ethyl acetate, reflux,
2 2
◦
h, 26%; (iii) mCPBA, acetic acid, RT, 6 h, 51%; (iv) Sn Me , Pd(PPh ) , dry toluene, 100 C, microwave,
2 6 3 4
4
4
◦
h, 51%; (v) Pd(PPh ) , DMF, 120 C, 24 h, 31%.
3
4
2
.3.1. Photochemical Characterization of Azobenzene Derivatives 3 and 4
The UV/VIS spectra of the azobenzene-functionalized axitinib derivatives
3
and
4
show absorption
(Figure 6a),
spectra that are typical for azobenzenes (Figure 6) [16
,
46]. For the meta-linked derivative 3
a maximum shift of the PSS toward the Z-isomer can be achieved by irradiation at 365 nm. By NMR
spectroscopy, a Z-isomer content of 83% was determined in the PSS. In comparison, the para-linked
derivative
measured. Back switching to the E-isomers E-
70 nm ( ), respectively. In both cases, the photoconversion is almost quantitative (for reasons of clarity
4
(Figure 6b) can be switched at 385 nm and for the resulting PSS, an E/Z ratio 20/80 was
3
and E- is possible by irradiation at 435 nm ( ) or
4
3
4
4
not shown in Figure 6). Furthermore, half-lives of the azobenzene derivatives were determined by
◦
UV/VIS spectroscopy at 37 C in DMSO. For the meta-linked compound
3, a half-life of 43.1 h was
obtained after fitting the experimental data, while para-linked derivative
4 possesses a much shorter
half-life of 5.7 h (Figures S3 and S4).
2
.3.2. Photochemical Characterization of Diazocine Derivatives 5–7
Sulfur–diazocine (linked in meta-position of the azo bond) shows a typical switching behavior
for diazocines (Figure 7a) [24 33]. After irradiation at 405 nm, a PSS with an E content of 47% can
be achieved. For para-linked derivative , only 25% of the E-isomer can be enriched by irradiation
at 405 nm. This low photoconversion can be explained by the poorly separated absorption bands
Figure 7b) that may originate from an electronic coupling of the diazocine scaffold with the -system
of the indazole. In contrast, the ethylene-bridged diazocine shows a decent photoconversion leading
5
,
6
(
π
7
to 60% of the E-isomer in the PSS (Figure 8). This trend correlates with the associated unsubstituted
diazocines, where a slightly better switching efficiency is also observed for the carbon-based system.

Int. J. Mol. Sci. 2020, 21, 8961
11 of 20
However, for all three derivatives
5
–7, the determined photoconversions are significantly lower than
for the corresponding unsubstituted diazocines, which can be switched to 70% (sulfur–diazocine) and
0% (carbon–diazocine) of the E-isomer, respectively [24,33].
9
a
b
Figure 6.
UV/VIS spectra of azobenzene-functionalized axitinib derivatives 3 and 4.
(
a
) UV/VIS-spectrum of meta-linked E-
after irradiation at 365 nm (blue). ( ) UV/VIS-spectrum of para-linked E-
as well as PSS after irradiation at 385 nm (blue). Spectra after irradiation at 435 nm (
) respectively are not shown for reasons of clarity. These spectra overlap with the spectra of the pure
3
(red) in DMSO (20
µ
M) as well as photostationary states (PSS)
(red) in DMSO (20 M)
) and 470 nm
b
4
µ
3
(
4
E-Isomers.
a
b
Figure 7. UV/VIS spectra of S-diazocine-functionalized axitinib derivatives. (
band of meta-linked Z- (red) in DMSO (100 M) as well as photostationary states (PSS) after irradiation
at 405 nm (blue). ( ) Enlarged view of n * band of para-linked Z- (red) in DMSO (100 M) as well as
a) Enlarged view of nπ*
5
µ
b
π
6
µ
PSS after irradiation at 405 nm (blue). Spectra of PSS after irradiation at 530 nm completely overlap
with the spectrum of the pure Z-isomers (not shown for reasons of clarity).
By irradiation at 530 nm, the diazocine-based compounds 5–7 can be switched back quantitatively
to the Z-isomer. Since the associated spectra completely overlap with the spectra of the pure Z-isomers,
they are not shown for reasons of clarity. To check the reversibility of the photoswitching, long-term
stability measurements with 20 switching cycles were performed. For this purpose, the compounds
were irradiated alternately for 20 s at 405 nm and 530 nm, respectively. After each irradiation,
an UV/VIS spectrum was recorded. After 20 switching cycles, none of the three derivatives 5–7 showed
significant photofatigue.

Int. J. Mol. Sci. 2020, 21, 8961
12 of 20
Figure 8. UV/VIS spectrum of C-diazocine-functionalized axitinib derivative (7). Enlarged view of nπ*
band of meta-linked Z-7 (red) in DMSO (100 µM) as well as photostationary states (PSS) after irradiation
at 405 nm (blue). Spectrum of PSS after irradiation at 530 nm is identical to the spectrum of the pure
Z-isomer (not shown for reasons of clarity).
◦
Half-lives at 37 C in DMSO were determined as 7.3 h for the meta-linked sulfur–diazocine
5
and
3.7 h for the para-linked S-diazocine
6
, whereas carbon–diazocine
7
exhibits a comparatively short
half-life of 1.5 h (Figures S5–S7). In summary, all compounds showed sufficient half-lives for in vitro
kinase testing.
It is noteworthy that the recording of UV/VIS-spectra in aqueous solution failed for the
diazocine-functionalized axitinib derivatives
detection limit.
5–7 due to compound precipitation above the minimum
2
.4. Biological Evaluation
2
.4.1. VEGFR-2 Kinase Assays
In a next step, the inhibitory activity of all photoresponsive axitinib derivatives against VEGFR-2
was tested using a luminescence-based ADP-Glo
reported before [14].
™ kinase assay under controlled light conditions as
VEGFR-2 Kinase Assays of Azobenzene Derivatives 3 and 4
Dose–response curves of derivatives and (Figure 9) show the moderate potency of both
3
4
irradiated and unirradiated compounds with IC50 values around 1000 nM. Overall, no distinct
difference in the biological activity of both azobenzene derivatives before and after irradiation could
be demonstrated. These findings are consistent with the modeling results, where plausible binding
modes were found for both isomers. However, for the two highest concentrations of both azobenzenes,
the kinase inhibition seems to be slightly stronger after irradiation (PSS 365 nm and 385 nm respectively)
than before irradiation. If one considers the high level of the plateaus in the dose–response curves
of the E-isomers, this effect might be caused by a limiting solubility of the E-isomer compared to the
Z-isomer.
VEGFR-2 Kinase Assay of Diazocine Derivatives 5–7
For the sulfur–diazocine compounds, the Z-isomers (illustrated in red) do not have any impact
on VEGFR-2 activity up to compound concentrations of 10,000 nM (Figure 10). In contrast, for the
irradiated compounds (PSS 405 nm, in blue), a significant decrease of the kinase activity with IC50
values of 214 nM for the meta-substituted diazocine (
respectively could be demonstrated. Considering that the IC50 values of the Z-isomers are higher than
0,000 nM, a difference in the biological activity before and after irradiation of at least a factor of 47 and
5) and 251 nM for the para-substituted diazocine (6)
1

Int. J. Mol. Sci. 2020, 21, 8961
13 of 20
40 can be observed respectively. These phototherapeutic indices are the highest differences reported
for reversibly switchable small molecule kinase inhibitors so far [10
–
12]. Beyond that, it has to be taken
) of the bioactive E-isomers,
are presumably quite potent with
IC50 values in the low nanomolar range. As a technical note, due to the poor water solubility of the
diazocine-functionalized axitinib derivatives , the kinase assay had to be performed with a final
into account that the present PSS contain only about 47% (
5) or 25% (6
respectively. Hence, we assume that the isolated isomers E- and E-
5
6
5–7
DMSO concentration of 10%. It is moteworthy that we confirmed that this high DMSO concentration
in the final kinase assay mixture does not affect VEGFR-2 kinase activity.
a
b
Figure 9. Dose–response curves of azobenzene-functionalized axitinib derivatives 3 and 4 in an in vitro
VEGFR-2 kinase assay with and without irradiation. Data points are means of double determinations
with their standard deviation as error bars. (
a
) E-
3
in the dark (red, IC50 = 1077 nM, residual activity
2
(
(
4
plateau): 33%, R = 0.90) as well as after irradiation at 365 nm (blue, IC = 1289 nM, residual activity
5
0
2
plateau): 5%, R = 0.90). (
3%, R = 0.95) as well as after irradiation at 385 nm (blue, IC50 = 1435 nM, residual activity (plateau):
b
) E-
4
in the dark (red, IC50 = 1020 nM, residual activity (plateau):
2
2
1
3%, R = 0.93).
a
b
Figure 10. Dose–response curves of sulfur–diazocine-functionalized axitinib derivatives
an in vitro VEGFR-2 kinase assay with and without irradiation. Data points are means of double
determinations with their standard deviation as error bars. Final DMSO concentration 10%. ( ) Z- in
5 and 6 in
a
5
the dark (red, IC50 > 10,000 nM) as well as after irradiation at 405 nm (blue, IC = 214 nM, residual
50
2
activity (plateau): 6%, R = 0.94). (
b) Z-6
in the dark (red, IC50 > 10,000 nM) as well as after irradiation
2
at 405 nm (blue, IC50 = 251 nM, residual activity (plateau): 10%, R = 0.94).
Dose–response analysis of the carbon–diazocine derivative
Z-isomer (Figure 11, red circles, residual activity: 37%) as well, which correlates with the results of
the induced-fit docking (Table 2). Therefore, the activation effect after irradiation is comparatively low
PSS 405 nm, IC50 = 493 nM). While the PSS of compound possesses the highest E ratio (60%), only a
weak inhibition of the kinase can be observed compared to the sulfur–diazocines 5 and 6.
7 reveals a weak activity for the
≥
(
7

Int. J. Mol. Sci. 2020, 21, 8961
14 of 20
Figure 11. Dose–response curves of carbon–diazocine-functionalized axitinib derivative 7 in an in vitro
VEGFR-2 kinase assay with and without irradiation. Data points are means of double determinations
with their standard deviation as error bars. Final DMSO concentration 10%. Z-
7
in the dark (red,
2
residual activity: ≥ 37%) as well as after irradiation at 405 nm (blue, IC50 = 493 nM, R = 0.89).
.4.2. Kinome Profiling (PamGene) of Sulfur–diazocine Derivative 5
2
Since sulfur–diazocine
this compound on a kinase profiling panel (PamGene [47
5
showed the best efficacy in the VEGFR-2 kinase assay, we further tested
49]) using lysates of human umbilical
–
vein endothelial cells (HUVECs). By using the PamGene technology, kinase activities are studied
via the phosphorylation degree of over 144 substrate peptides fixed in arrays. In our testing,
the profiling reveals strong kinase inhibition after treatment with E-axitinib (
1) compared to DMSO
(negative control). This result is in agreement with our findings in the previous axitinib study [14].
In contrast, HUVEC lysate treated with unirradiated derivative
5 shows similar kinase activities as the
DMSO control, demonstrating the Z-isomer to be biologically inactive (Figure 12). Irradiation compound
leads to a moderate decrease of kinase activities, indicating a modest cellular effectivity of the PSS of
diazocine with a trend toward the positive control E-axitinib ( ). As described above, 10% DMSO
5
5
1
as co-solvent was also necessary in this assay to avoid compound precipitation of the micromolar
compound solution.
Figure 12. Kinome activity profiling of treated human umbilical vein endothelial cell (HUVEC) cell
lysate. Heat map shows the phosphorylation degree of affected peptides from red (high phosphorylation
degree) to blue (low phosphorylation degree). A lower phosphorylation degree confirms more inhibited
protein kinases. Lysates were treated with 20
µM diazocine-functionalized axitinib derivative 5 with or
without irradiation at 405 nm, 1 µM axitinib or 10% DMSO as negative control.

Int. J. Mol. Sci. 2020, 21, 8961
15 of 20
3
. Discussion
In this study, a small set of azobenzene- and diazocine-functionalized derivatives of the approved
VEGFR kinase inhibitor axitinib was designed, synthesized, and characterized. For sulfur–diazocine
derivatives and , irradiation induced significant biological effects in an in vitro VEGFR-2 kinase
assay. In agreement with the data of the molecular modeling studies, the Z-isomers of diazocine
derivatives and are biologically inactive, while strong kinase inhibition could be measured for
the PSS after irradiation at 405 nm, resulting in IC50 values of 214 nM ( ) and 251 nM ( ) respectively.
5
6
5
6
5
6
Hence, phototherapeutic indices of 47 and 40 are attained, with regard to the difference between the
non-irradiated and irradiated compounds. This is a major improvement compared to previously
reported photoswitchable kinase inhibitors with 1.6- to 11-fold differences [10
a modest difference in the phosphorylation profiles of HUVEC cell lysates with and without irradiation
could be demonstrated for compound . Remarkably, for the carbon–diazocine , the Z-isomer also
–12]. Furthermore,
5
7
shows moderate kinase inhibition. Modeling data suggest that this result could be explained by the
larger angle between both phenyl rings of the carbon–diazocine in the Z-configuration.
For the azobenzene derivatives
3 and 4, both molecular modeling and biological testing revealed
no significant differences in the binding affinities of the E/Z-isomers. Regarding biological applications
of photoresponsive compounds, these results suggest diazocines to have advantages over standard
azobenzenes: Due to large changes of the molecular geometry upon irradiation, diazocines are
promising photoswitches compared to the less rigid azobenzenes, where the rotation of a C–N single
bond can weaken photoinduced effects toward ligand–target binding. However, the physicochemical
parameters of the diazocine-functionalized compounds such as poor water solubility have to be
optimized. In this context, Lentes et al., published e.g., a nitrogen-bridged diazocine that shows
great switching properties as well as improved water solubility [50]. Thus, a key challenge for the
development of photoresponsive diazocines toward biological applications will be to focus on both the
optimization of photochemical properties and physicochemical parameters.
4
. Materials and Methods
4
.1. Computational Chemistry
For molecular modeling studies, the software Maestro (v.11.7, Release 2018-03), Schrödinger
LLC (New York, NY, USA) was used. Calculations were run on a DELL Precision T3610 computer
Round Rock, Texas, USA). The protein structure 4AG8 of the RCSB protein data bank (PDB) was used
(
as a VEGFR-2 model. The protein structure was prepared with the Protein Preparation Wizard prior to
docking. Bond orders were adjusted, hydrogen atoms were added, disulfide bonds were optimized,
and water molecules within a distance > 5 Å to heteroatoms were deleted. Missing residues and
missing loops were added using the tool Prime. H-bonding within the protein structure was optimized
using the standard protocol in Glide. The geometry of the protein was improved in a simplified,
restricted optimization using an OPLS3e force field. In the process, heavy atoms within an RMSD of
0
.3 Å were converged. Receptor Grids were created using the tool Glide.
Geometries of E- and Z-azobenzene moieties as well as all diazocines structures were optimized
using DFT on B3LYP/6-31G* level of theory. Binding modes were calculated with the tool Glide
using Extra Precision (XP) mode. The options Canonicalize input conformation and post-docking
minimization were deactivated to prevent conformational changes of the quantum-mechanically
optimized structures. In addition, quantum-mechanically optimized diazocine- and azobenzene
moieties were frozen during Glide dockings using torsional constraints. For dockings into the VEGFR-2
crystal structure (PDB: 4AG8), H-bond constraints to the residues Cys-919, Asp-1046, Glu-917 und
Glu-885 were defined. Induced-fit binding modes were calculated with the tool Induced-fit-docking
(IFD) using the standard protocol and OPLS3e force field. Torsional constraints for diazocine- and
azobenzene moieties during IFD dockings were defined programmatically.

Int. J. Mol. Sci. 2020, 21, 8961
16 of 20
4
.2. Synthesis
Detailed synthesis descriptions and analytical data of all compounds can be found in the
1
Supplementary Materials. NMR spectra were recorded either on a Bruker Avance III 300 ( H-NMR:
15
119
3
(
6
00 MHz, 13C-NMR: 75 MHz, N-NMR: 30 MHz,
Sn-NMR: 112 MHz), a Bruker Ascend 400
H-NMR: 400 MHz, C-NMR: 100 MHz, 19F-NMR: 377 MHz, 119Sn-NMR: 149 MHz), or a Bruker AV
1
13
1
13
00 ( H-NMR: 600 MHz, C-NMR: 150 MHz) (Bruker BioSpin, Rheinstetten, Germany). The spectra are
referenced to the residual signals of the deuterated solvents: Aceton-d : 2.05 ppm ( H-NMR), 29.84 ppm
C-NMR); Chloroform: 7.26 ppm ( H-NMR), 77.16 ppm ( C-NMR); DMSO-d : 2.50 ppm ( H-NMR),
6
9.52 ppm ( C-NMR); Methanol-d : 3.31 ppm ( H-NMR), 49.0 ppm ( C-NMR). NMR signals were
1
6
1
3
1
13
1
(
13
1
13
3
4
analyzed using the following abbreviations: singlet (s), broad singlet (bs), doublet (d), doublet of
doublets (dd), doublet of doublets of doublets (ddd), doublet of triplets (dt), triplet (t), triplet of doublets
(
td), quartet (q), multiplet (m), centered multiplet (mc.). For explicit assignments of signals 2D NMR
1
13
spectra (COSY, HSQC, HMBC) were used. The actual H- and C-NMR spectra of key compounds
(azobenzene derivatives 3 and 4 as well as diazocine derivatives 5–7) are shown in Figures S10–S19.
LC-MS spectra were recorded on a Bruker Esquire ~LC ion trap mass spectrometer (Bremen,
◦
Germany) in the positive ion mode (dry gas 6.5 L/min, nebulizer 25 psi, drying temperature 250 C)
after chromatographic separation using an Agilent 1100 HPLC system (Waldbronn, Germany) with
an RP-8 column (Waters Xterra MS C8, 50
×
4.6 mm, 3.5 µm, Milford, MA, USA) and a 0.1% acetic
acid/acetonitrile gradient. High-resolution mass spectra (HRMS) were recorded on a JEOL AccuTOF
GCv 4G electron ionization time of flight (EI-TOF) mass spectrometer (Freising, Germany). ESI mass
spectra were recorded on a ThermoFisher Q Exactive Plus Hybrid Quadrupol-Orbitrap spectrometer
(Dreieich, Germany).
HPLC analysis was performed on a Hewlett-Packard 1050 Series system (Palo Alto, CA, USA)
using a ZORBAX Eclipse XDB-C8 column (Agilent, Santa Clara, CA, USA).
4
.3. Photochemical Characterization
UV/VIS spectra were measured using a Varian Cary® 50 Scan UV/VIS photometer from Agilent
◦
◦
(
Waldbronn, Germany) equipped with a Varian Cary PCB 150 thermostat either at 25 C or 37 C.
For irradiation experiments, the following custom-made light sources have been used: 365 nm
LED reactor (Sahlmann Photochemical Solutions, Bad Segeberg, Germany, 12 450 mW Nichia
NCSU033B LEDs, dimmable to 25%, 50%, 75% or 100%), 385 nm LED lamp (Sahlmann Photochemical
Solutions, 3 2430 mW Nichia NC4U134 LEDs, not dimmable), 405 nm LED lamp (Sahlmann
Photochemical Solutions, 3 2430 mW Nichia NVSU233A-U405 LEDs, not dimmable), 420 nm LED
lamp (Sahlmann Photochemical Solutions, 1 2300 mW Marubeni 420-66-60 LED, not dimmable),
30 nm LED lamp (Sahlmann Photochemical Solutions, 8 575 mW Nichia NCSG219-V1 LEDs,
×
×
×
×
5
×
dimmable to 25%, 50%, 75% or 100%). The fitting of thermal relaxations was done with GraphPad
®
Prism (v.7.03, GraphPad Software, San Diego, CA, USA).
4
.4. Kinase Assays
Kinase activity of VEGFR-2 (purified kinase domain, KDR active, Promega) was determined
using a luminescence based ADP-Glo
™ kinase assay (Promega, Madison, WI, USA). In this assay,
ADP generated during the kinase reaction is transformed into a luminescence signal, which is then
proportional to the kinase activity. The assay was performed according to the manufacturer’s protocol.
Briefly, white 96-well CulturePlates
reaction mixtures of total 25 L. Kinase reactions were run with 10
.25 mg/mL substrate (Poly(4:1 Glu, Tyr)Peptide) in kinase buffer (40 mM Tris (pH 7.5), 20 mM MgCl2,
.1 mg/mL BSA, 2 mM DTT, 2 mM MnCl and 100 M Na VO ). Test compounds were dissolved in
™
(PerkinElmer, Waltham, MA, USA) were used with kinase
µ
µ
M ATP, 0.4 ng/ L kinase and
µ
0
0
µ
2
3
4
DMSO, and 1:3 serial dilutions were prepared. As described previously [14], compound handling took
place under controlled light conditions. Since it was not possible to examine switching behavior of the

Int. J. Mol. Sci. 2020, 21, 8961
17 of 20
axitinib derivatives in aqueous solution, compound irradiation was first performed in DMSO before
adding this solution to the aqueous kinase assay mixture. Azobenzene was irradiated at 365 nm using
a dimmable custom-made lamp for 96-well plates, consisting of 16 LEDs in the distance of the wells
Sahlmann Photochemical Solutions, 16 750 mW Nichia NCSU276A LEDs). Every well was irradiated
for 20 s at 5% power. Azobenzene was irradiated at 385 nm using a hand lamp consisting of three
LEDs (Sahlmann Photochemical Solutions, 3 2430 mW Nichia NC4U134 LEDs). Here, the whole plate
was irradiated for 2 min. Diazocines were irradiated at 405 nm using a custom-made 96-well plate
lamp (Sahlmann Photochemical Solutions, 16 980 mW Nichia NVSU233A-U405 LEDs, dimmable).
Every well was irradiated for 20 s at 5% power. Diazocines were additionally irradiated every 30 min
during the assay procedure due to the short half-life of compound . For diazocine testing, the 405 nm
lamp was used as the only light source. Before kinase addition starts the reaction, compounds were
added to the reaction mixture resulting in final concentrations of 1% (azo-compounds ) or 10%
diazocines ) DMSO and inhibitor concentrations ranging from 10 M to 0.5 nM. It was verified that
0% DMSO in the final kinase assay mixture does not affect VEGFR-2 kinase activity. Luminescence
3
(
×
4
×
5–7
×
5–7
7
2–4
(
5
–7
µ
1
®
intensities were recorded on a FLUOstar Omega (BMG Labtech, Ortenberg, Germany). Kinase activity
was calculated in percent of control without inhibitor (DMSO only) and plotted against logarithm of
inhibitor concentration. Data points are means of double measurements with standard deviation as
error bars. Sigmoidal fitting (log(inhibitor) vs. response–variable slope) and calculation of IC50 values
®
was performed using GraphPad Prism (v.7.03).
4
.5. Kinase Selectivity Profiling (PamGene)
Human Umbilical Vein Endothelial Cells (HUVECs) were obtained from PromoCell (Heidelberg,
◦
Germany) and cultivated with ECGM2 (PromoCell) at 37 C and 5% CO in humidified atmosphere.
2
2
For kinome profiling, HUVECs were seeded in cell culture flasks (25 cm ) to 90% confluence. After cell
lysis with M-Per and centrifugation at 4 C, cell lysates were immediately stored at
Protein concentration was determined with Bradford assay (ThermoFisher, Waltham, MA, USA).
Samples for profiling were prepared according to manufacturer’s protocol. For kinome profiling,
PamStation 12 was used (PamGene, ’s-Hertogenbosch, Netherlands). Protein tyrosine kinase
PamChips® were used with a total of 5
µg protein and 100 µM ATP per array. The microchip
◦
◦
−
80 C [14].
®
for kinome profiling contains 144 peptides linked to a porous surface. These peptides can be
phosphorylated by cell lysate kinases. Readout was achieved by washing microchips with fluorescent
antibodies. Compounds in DMSO as well as lysate master mix (including compounds) were irradiated
at 405 nm (irradiation of reaction tubes with 405 nm well plate lamp, Sahlmann Photochemical Solutions,
1
6 × 980 mW Nichia NVSU233A-U405 LEDs, 20% power, for 30 s), while compound handling was
established under controlled lighting. Profiling was performed at a final DMSO concentration of 10%.
Data analysis was performed as previously described by Labots et al. [51].
Supplementary Materials: Supplementary materials can be found at http://www.mdpi.com/1422-0067/21/23/
8
961/s1.
Author Contributions: Conceptualization, D.S., L.H., C.P. and R.H.; methodology, D.S., L.H., T.R., L.W. and
J.E.; software, D.S.; validation, L.H. and T.R.; formal analysis, D.S., L.H., T.R., L.W. and J.E.; resources, C.P.,
R.H. and M.K.; data curation, D.S., L.H., T.R. and L.W.; writing—original draft preparation, D.S., L.H. and T.R.;
writing—review and editing, C.P., L.W. and R.H.; visualization, D.S., L.H. and T.R.; supervision, C.P., R.H. and
M.K.; project administration, D.S.; funding acquisition, C.P. and R.H. All authors have read and agreed to the
published version of the manuscript.
Funding: This research was funded by DFG (German Research Society) grant Collaborative Research Centre 677
“
Function by Switching” and DFG grant PE1605_2_2.
Acknowledgments: We thank Martin Schütt (Institute of Pharmacy, Kiel University) and Konstantin Hoffer
University Medical Center Hamburg) for excellent technical assistance and Fynn Röhricht (Otto-Diels-Institute
(
of Organic Chemistry, Kiel University) for providing quantum mechanically optimized diazocine structures.
We gratefully acknowledge the support of Dieter Schollmeyer (Institute for Organic Chemistry, University of
Mainz, Germany) for X-ray analysis of small molecule compounds.

Int. J. Mol. Sci. 2020, 21, 8961
18 of 20
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the
study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to
publish the results.
Abbreviations
BSA
Bovine serum albumin
Dibenzoyl peroxide
Dichloromethane
Dimethylformamide
Dimethyl sulfoxide
DBPO
DCM
DMF
DMSO
DTT
Dithiothreitol
FDA
HUVEC
LED
NBS
pdb
Food and Drug Administration
Human umbilical vein endothelial cell
Light-emitting diode
N-Bromosuccinimide
Protein Data Bank
PKC
Protein kinase C
PSS
RMSD
RT
Photostationary state
Root mean square deviation
Room temperature
THF
Tetrahydrofuran
VEGFR
Vascular endothelial growth factor receptor
References
1.
2.
3.
4.
5.
Velema, W.A.; Szymanski, W.; Feringa, B.L. Photopharmacology: Beyond proof of principle. J. Am. Chem. Soc.
014, 136, 2178–2191. [CrossRef] [PubMed]
Lerch, M.M.; Hansen, M.J.; van Dam, G.M.; Szymanski, W.; Feringa, B.L. Emerging Targets in
2
Photopharmacology. Angew. Chem. Int. Ed. Engl. 2016, 55, 10978–10999. [CrossRef] [PubMed]
Broichhagen, J.; Frank, J.A.; Trauner, D. A roadmap to success in photopharmacology. Acc. Chem. Res. 2015
8, 1947–1960. [CrossRef] [PubMed]
Hüll, K.; Morstein, J.; Trauner, D. In Vivo Photopharmacology. Chem. Rev. 2018, 118, 10710–10747. [CrossRef]
PubMed]
,
4
[
Szyma n´ ski, W.; Beierle, J.M.; Kistemaker, H.A.V.; Velema, W.A.; Feringa, B.L. Reversible photocontrol of
biological systems by the incorporation of molecular photoswitches. Chem. Rev. 2013, 113, 6114–6178.
[CrossRef] [PubMed]
6
7
8
9
1
1
.
.
.
.
Bournez, C.; Carles, F.; Peyrat, G.; Aci-Sèche, S.; Bourg, S.; Meyer, C.; Bonnet, P. Comparative Assessment of
Protein Kinase Inhibitors in Public Databases and in PKIDB. Molecules 2020, 25, 3226. [CrossRef] [PubMed]
Carles, F.; Bourg, S.; Meyer, C.; Bonnet, P. PKIDB: A Curated, Annotated and Updated Database of Protein
Kinase Inhibitors in Clinical Trials. Molecules 2018, 23, 908. [CrossRef]
Roskoski, R. Properties of FDA-approved small molecule protein kinase inhibitors: A 2020 update.
Pharmacol. Res. 2020, 152, 104609. [CrossRef]
Fleming, C.L.; Grøtli, M.; Andr
ChemPhotoChem 2019, 3, 318–326. [CrossRef]
0. Ferreira, R.; Nilsson, J.R.; Solano, C.; Andr asson, J.; Grøtli, M. Design, Synthesis and Inhibitory Activity of
éasson, J. On-Command Regulation of Kinase Activity using Photonic Stimuli.
é
Photoswitchable RET Kinase Inhibitors. Sci. Rep. 2015, 5, 9769. [CrossRef]
1. Schehr, M.; Ianes, C.; Weisner, J.; Heintze, L.; Müller, M.P.; Pichlo, C.; Charl, J.; Brunstein, E.; Ewert, J.;
Lehr, M.; et al. 2-Azo-, 2-diazocine-thiazols and 2-azo-imidazoles as photoswitchable kinase inhibitors:
Limitations and pitfalls of the photoswitchable inhibitor approach. Photochem. Photobiol. Sci. 2019
8, 1398–1407. [CrossRef] [PubMed]
,
1
1
2. Hoorens, M.W.H.; Ourailidou, M.E.; Rodat, T.; van der Wouden, P.E.; Kobauri, P.; Kriegs, M.; Peifer, C.;
Feringa, B.L.; Dekker, F.J.; Szymanski, W. Light-controlled inhibition of BRAFV600E kinase. Eur. J. Med. Chem.
2
019, 179, 133–146. [CrossRef] [PubMed]

Int. J. Mol. Sci. 2020, 21, 8961
19 of 20
1
1
3. Wilson, D.; Li, J.W.; Branda, N.R. Visible-Light-Triggered Activation of a Protein Kinase Inhibitor.
ChemMedChem 2017, 12, 284–287. [CrossRef] [PubMed]
4. Schmidt, D.; Rodat, T.; Heintze, L.; Weber, J.; Horbert, R.; Girreser, U.; Raeker, T.; Bußmann, L.; Kriegs, M.;
Hartke, B.; et al. Axitinib: A Photoswitchable Approved Tyrosine Kinase Inhibitor. ChemMedChem 2018
3, 2415–2426. [CrossRef] [PubMed]
5. Mittal, K.; Wood, L.S.; Rini, B.I. Axitinib in Metastatic Renal Cell Carcinoma. Biol. Ther. 2012, 2, 5. [CrossRef]
PubMed]
6. Beharry, A.A.; Woolley, G.A. Azobenzene photoswitches for biomolecules. Chem. Soc. Rev. 2011, 40, 4422–4437.
CrossRef]
7. Wei, N.; Liang, J.; Peng, S.; Sun, Q.; Dai, Q.; Dong, M. Design, Synthesis, and Biological Evaluation of Axitinib
Derivatives. Molecules 2018, 23, 747. [CrossRef]
8. Weston, C.E.; Richardson, R.D.; Fuchter, M.J. Photoswitchable basicity through the use of azoheteroarenes.
Chem. Commun. (Camb) 2016, 52, 4521–4524. [CrossRef]
9. Leippe, P. Tethered Photopharmacology. Ph.D. Thesis, Ludwig-Maximilians-Universität München, Munich,
Germany, 2018.
0. Simeth, N.A.; Crespi, S.; Fagnoni, M.; König, B. Tuning the Thermal Isomerization of Phenylazoindole
Photoswitches from Days to Nanoseconds. J. Am. Chem. Soc. 2018, 140, 2940–2946. [CrossRef]
1. Otsuki, J.; Suwa, K.; Sarker, K.K.; Sinha, C. Photoisomerization and thermal isomerization of
arylazoimidazoles. J. Phys. Chem. A 2007, 111, 1403–1409. [CrossRef]
,
1
1
1
1
1
1
2
2
2
[
[
2. Crespi, S.; Simeth, N.A.; Bellisario, A.; Fagnoni, M.; König, B. Unraveling the Thermal Isomerization
Mechanisms of Heteroaryl Azoswitches: Phenylazoindoles as Case Study. J. Phys. Chem.
23, 1814–1823. [CrossRef] [PubMed]
3. Rustler, K.; Nitschke, P.; Zahnbrecher, S.; Zach, J.; Crespi, S.; König, B. Photochromic Evaluation of
(5)-Arylazo-1H-pyrazoles. J. Org. Chem. 2020, 85, 4079–4088. [CrossRef] [PubMed]
4. Siewertsen, R.; Neumann, H.; Buchheim-Stehn, B.; Herges, R.; Näther, C.; Renth, F.; Temps, F. Highly efficient
reversible Z-E photoisomerization of a bridged azobenzene with visible light through resolved S(1)(n pi*)
absorption bands. J. Am. Chem. Soc. 2009, 131, 15594–15595. [CrossRef]
A 2019,
1
2
2
3
25. Samanta, S.; Qin, C.; Lough, A.J.; Woolley, G.A. Bidirectional photocontrol of peptide conformation with a
bridged azobenzene derivative. Angew. Chem. Int. Ed. Engl. 2012, 51, 6452–6455. [CrossRef] [PubMed]
26. Eljabu, F.; Dhruval, J.; Yan, H. Incorporation of cyclic azobenzene into oligodeoxynucleotides for the
photo-regulation of DNA hybridization. Bioorg. Med. Chem. Lett. 2015, 25, 5594–5596. [CrossRef]
27. Trads, J.B.; Hüll, K.; Matsuura, B.S.; Laprell, L.; Fehrentz, T.; Görldt, N.; Kozek, K.A.; Weaver, C.D.;
Klöcker, N.; Barber, D.M.; et al. Sign Inversion in Photopharmacology: Incorporation of Cyclic Azobenzenes
in Photoswitchable Potassium Channel Blockers and Openers. Angew. Chem. Int. Ed. Engl. 2019
8, 15421–15428. [CrossRef]
,
5
2
8. Thapaliya, E.R.; Zhao, J.; Ellis-Davies, G.C.R. Locked-Azobenzene: Testing the Scope of a Unique
Photoswitchable Scaffold for Cell Physiology. ACS Chem. Neurosci. 2019, 10, 2481–2488. [CrossRef]
9. McTigue, M.; Murray, B.W.; Chen, J.H.; Deng, Y.-L.; Solowiej, J.; Kania, R.S. Molecular conformations,
interactions, and properties associated with drug efficiency and clinical performance among VEGFR TK
inhibitors. Proc. Natl. Acad. Sci. USA 2012, 109, 18281–18289. [CrossRef]
2
3
3
3
3
0. Friesner, R.A.; Banks, J.L.; Murphy, R.B.; Halgren, T.A.; Klicic, J.J.; Mainz, D.T.; Repasky, M.P.; Knoll, E.H.;
Shelley, M.; Perry, J.K.; et al. Glide: A new approach for rapid, accurate docking and scoring. 1. Method and
assessment of docking accuracy. J. Med. Chem. 2004, 47, 1739–1749. [CrossRef]
1. Halgren, T.A.; Murphy, R.B.; Friesner, R.A.; Beard, H.S.; Frye, L.L.; Pollard, W.T.; Banks, J.L. Glide: A new
approach for rapid, accurate docking and scoring. 2. Enrichment factors in database screening. J. Med. Chem.
2
004, 47, 1750–1759. [CrossRef]
2. Friesner, R.A.; Murphy, R.B.; Repasky, M.P.; Frye, L.L.; Greenwood, J.R.; Halgren, T.A.; Sanschagrin, P.C.;
Mainz, D.T. Extra precision glide: Docking and scoring incorporating a model of hydrophobic enclosure for
protein-ligand complexes. J. Med. Chem. 2006, 49, 6177–6196. [CrossRef] [PubMed]
3. Hammerich, M.; Schütt, C.; Stähler, C.; Lentes, P.; Röhricht, F.; Höppner, R.; Herges, R. Heterodiazocines:
Synthesis and Photochromic Properties, Trans to Cis Switching within the Bio-optical Window. J. Am.
Chem. Soc. 2016, 138, 13111–13114. [CrossRef] [PubMed]

Int. J. Mol. Sci. 2020, 21, 8961
20 of 20
3
3
4. Sherman, W.; Day, T.; Jacobson, M.P.; Friesner, R.A.; Farid, R. Novel procedure for modeling ligand/receptor
induced fit effects. J. Med. Chem. 2006, 49, 534–553. [CrossRef] [PubMed]
5. Chekal, B.P.; Guinness, S.M.; Lillie, B.M.; McLaughlin, R.W.; Palmer, C.W.; Post, R.J.; Sieser, J.E.; Singer, R.A.;
Sluggett, G.W.; Vaidyanathan, R.; et al. Development of an Efficient Pd-Catalyzed Coupling Process for
Axitinib. Org. Process Res. Dev. 2013, 18, 266–274. [CrossRef]
3
3
3
3
4
4
4
4
4
4
6. Zhai, L.-H.; Guo, L.-H.; Luo, Y.-H.; Ling, Y.; Sun, B.-W. Effective Laboratory-Scale Preparation of Axitinib by
Two CuI-Catalyzed Coupling Reactions. Org. Process Res. Dev. 2015, 19, 849–857. [CrossRef]
7. Correa, A.; Tellitu, I.; Dom
í
nguez, E.; SanMartin, R. Novel Alternative for the N
−
S Bond Formation and Its
Application to the Synthesis of Benzisothiazol-3-ones. Org. Lett. 2006, 8, 4811–4813. [CrossRef]
8. Sridhara, M.B.; Srinivasa, G.R.; Channe Gowda, D. Ammonium chloride mediated reduction of azo
compounds to hydrazo compounds. J. Chem. Res. 2004, 74–75. [CrossRef]
9. Ko z´ lecki, T.; Syper, L.; Wilk, K.A. 4-Lithio-4’-alkylazobenzenes as Convenient Intermediates for the
Preparation of Azobenzene Derivatives. Synthesis 1997, 681–684. [CrossRef]
0. Strueben, J.; Gates, P.J.; Staubitz, A. Tin-functionalized azobenzenes as nucleophiles in Stille cross-coupling
reactions. J. Org. Chem. 2014, 79, 1719–1728. [CrossRef]
1. Strueben, J.; Lipfert, M.; Springer, J.-O.; Gould, C.A.; Gates, P.J.; Sönnichsen, F.D.; Staubitz, A.
High-yield lithiation of azobenzenes by tin-lithium exchange. Chemistry 2015, 21, 11165–11173. [CrossRef]
2. Schehr, M.; Hugenbusch, D.; Moje, T.; Näther, C.; Herges, R. Synthesis of mono-functionalized S-diazocines
via intramolecular Baeyer–Mills reactions. Beilstein J. Org. Chem. 2018, 14, 2799–2804. [CrossRef] [PubMed]
3. Moormann, W.; Langbehn, D.; Herges, R. Synthesis of functionalized diazocines for application as building
blocks in photo- and mechanoresponsive materials. Beilstein J. Org. Chem. 2019, 15, 727–732. [CrossRef]
4. Moormann, W.; Langbehn, D.; Herges, R. Solvent-Free Synthesis of Diazocine. Synthesis 2017, 49, 3471–3475.
[CrossRef]
5. Maier, M.S.; Hüll, K.; Reynders, M.; Matsuura, B.S.; Leippe, P.; Ko, T.; Schäffer, L.; Trauner, D. Oxidative
Approach Enables Efficient Access to Cyclic Azobenzenes. J. Am. Chem. Soc. 2019, 141, 17295–17304.
[
CrossRef]
6. Bl ger, D.; Hecht, S. Visible-Light-Activated Molecular Switches. Angew. Chem. Int. Ed. Engl. 2015
4, 11338–11349. [CrossRef] [PubMed]
7. Tahiri, A.; Røe, K.; Ree, A.H.; de Wijn, R.; Risberg, K.; Busch, C.; Lønning, P.E.; Kristensen, V.; Geisler, J.
Differential inhibition of ex-vivo tumor kinase activity by vemurafenib in BRAF(V600E) and BRAF wild-type
metastatic malignant melanoma. PLoS ONE 2013, 8, e72692. [CrossRef] [PubMed]
4
4
é
,
5
4
8. Arni, S.; Le, T.H.N.; de Wijn, R.; Garcia-Villegas, R.; Dankers, M.; Weder, W.; Hillinger, S. Ex vivo multiplex
profiling of protein tyrosine kinase activities in early stages of human lung adenocarcinoma. Oncotarget 2017
,
8
, 68599–68613. [CrossRef] [PubMed]
4
5
9. Baharani, A.; Trost, B.; Kusalik, A.; Napper, S. Technological advances for interrogating the human kinome.
Biochem. Soc. Trans. 2017, 45, 65–77. [CrossRef]
0. Lentes, P.; Stadler, E.; Röhricht, F.; Brahms, A.; Gröbner, J.; Sönnichsen, F.D.; Gescheidt, G.; Herges, R.
Nitrogen Bridged Diazocines: Photochromes Switching within the Near-Infrared Region with High Quantum
Yields in Organic Solvents and in Water. J. Am. Chem. Soc. 2019, 141, 13592–13600. [CrossRef]
5
1. Labots, M.; Gotink, K.J.; Dekker, H.; Azijli, K.; van der Mijn, J.C.; Huijts, C.M.; Piersma, S.R.; Jiménez, C.R.;
Verheul, H.M.W. Evaluation of a tyrosine kinase peptide microarray for tyrosine kinase inhibitor therapy
selection in cancer. Exp. Mol. Med. 2016, 48, e279. [CrossRef]
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.
©
2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
CC BY) license (http://creativecommons.org/licenses/by/4.0/).
(