Covalent inhibition of histone deacetylase 8 by 3,4-dihydro-2H-pyrimido[1,2-c][1,3]benzothiazin-6-imine

Article abstract of DOI:10.1016/j.bbagen.2019.01.001

Background: HDAC8 is an established target for T-cell lymphoma and childhood neuroblastoma. Benzothiazine-imines are promising HDAC8 inhibitors with unknown binding mechanism lacking a usual zinc binding group. Methods: In this study high-resolution and quantitative HPLC-coupled ESI-MS/MS techniques are combined with crystal structure determination and a variety of biochemical and computational methods to elucidate the reaction mechanism between benzothiazine-imine 1 and HDAC8. Results: 1) 1 is a covalent inhibitor of HDAC8; 2) inhibition is reversible in the presence of reducing agents; 3) C153 in the active site and C102 are involved in the inhibition mechanism; 4) 1 modifies various cysteines in HDAC8 forming either thiocyanates or mixed disulfides with 3; 5) 1 and 5 dock in close proximity to C153 within the active site. This is supposed to accelerate covalent inactivation particularly in HDAC8 and suggested as major determinant for the observed nanomolar potency and selectivity of 1. Conclusions: 1 and its analogs are interesting model compounds but unsuitable for therapeutic treatment due to their high unselective reactivity towards thiol groups. However, the postulated preceding non-covalent binding mode of 1 opens a door to optimized next generation compounds that combine potent and selective non-covalent recognition with low reactivity towards C153 at the active site of HDAC8. General significance: 1 represents a completely new class of inhibitors for HDAC8. Initial non-covalent interaction at the bottom of the active site is suggested to be the key for its selectivity. Further optimization of non-covalent interaction and thiol-reactivity provides opportunities to develop therapeutic useful covalent HDAC8 inhibitors.

Full text of DOI:10.1016/j.bbagen.2019.01.001

Accepted Manuscript
Covalent inhibition of histone deacetylase 8 by 3,4-dihydro-2H-
pyrimido[1,2-c][1,3]benzothiazin-6-imine
Marius Muth, Niklas Jänsch, Aleksandra Kopranovic, Andreas
Krämer, Nathalie Wössner, Manfred Jung, Frank Kirschhöfer,
Gerald Brenner-Weiß, Franz-Josef Meyer-Almes
PII:
S0304-4165(19)30001-7
DOI:
Reference:
https://doi.org/10.1016/j.bbagen.2019.01.001
BBAGEN 29279
To appear in:
BBA - General Subjects
Received date:
Revised date:
Accepted date:
31 August 2018
25 November 2018
2 January 2019
Please cite this article as: Marius Muth, Niklas Jänsch, Aleksandra Kopranovic, Andreas
Krämer, Nathalie Wössner, Manfred Jung, Frank Kirschhöfer, Gerald Brenner-Weiß,
Franz-Josef Meyer-Almes , Covalent inhibition of histone deacetylase 8 by
3,4-dihydro-2H-pyrimido[1,2-c][1,3]benzothiazin-6-imine.
https://doi.org/10.1016/j.bbagen.2019.01.001
Bbagen
(2019),
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.

ACCEPTED MANUSCRIPT
Covalent inhibition of histone deacetylase 8 by
3,4-dihydro-2H-pyrimido[1,2-c][1,3]benzothiazin-6-imine
Marius Muth1,3, Niklas Jänsch1, Aleksandra Kopranovic1, Andreas Krämer2, Nathalie Wössner4,
Manfred Jung4, Frank Kirschhöfer3, Gerald Brenner-Weiß3, Franz-Josef Meyer-Almes1,*
1 Department of Chemical Engineering and Biotechnology, University of Applied Sciences
Darmstadt, Germany.
2 Institute of Pharmaceutical Chemistry, University Frankfurt, Germany.
3 Bioengineering and Biosystems, Institute of Functional Interfaces, Karlsruhe Institute of
Technology, Germany.
4 Institute of Pharmaceutical Sciences, University of Freiburg, Germany.
* Corresponding author:
Franz-Josef Meyer-Almes
Email : franz-josef.meyer-almes@h-da.de
Phone: +49 6151168406
Fax : +49 6151168404
1

ACCEPTED MANUSCRIPT
Background:
HDAC8 is an established target for T-cell lymphoma and childhood neuroblastoma.
Benzothiazine-imines are promising HDAC8 inhibitors with unknown binding mechanism
lacking a usual zinc binding group.
Methods:
In this study high-resolution and quantitative HPLC-coupled ESI-MS / MS techniques are
combined with crystal structure determination and a variety of biochemical and computational
methods to elucidate the reaction mechanism between benzothiazine-imine 1 and HDAC8.
Results:
1) 1 is a covalent inhibitor of HDAC8; 2) inhibition is reversible in the presence of reducing
agents; 3) C153 in the active site and C102 are involved in the inhibition mechanism; 4) 1
modifies various cysteines in HDAC8 forming either thiocyanates or mixed disulfides with 3; 5)
1 and 5 dock in close proximity to C153 within the active site. This is supposed to accelerate
covalent inactivation particularly in HDAC8 and suggested as major determinant for the
observed nanomolar potency and selectivity of 1.
Conclusions:
1 and its analogs are interesting model compounds but unsuitable for therapeutic treatment due to
their high unselective reactivity towards thiol groups. However, the postulated preceding non-
covalent binding mode of 1 opens a door to optimized next generation compounds that combine
potent and selective non-covalent recognition with low reactivity towards C153 at the active site
of HDAC8.
2

ACCEPTED MANUSCRIPT
General Significance:
1 represents a completely new class of inhibitors for HDAC8. Initial non-covalent interaction at
the bottom of the active site is suggested to be the key for its selectivity. Further optimization of
non-covalent interaction and thiol-reactivity provides opportunities to develop therapeutic useful
covalent HDAC8 inhibitors.
Keywords:
HDAC8, selective inhibition, binding mechanism, non-hydroxamate inhibitor, covalent inhibitor,
enzyme inhibition
3

ACCEPTED MANUSCRIPT
1. Introduction
3,4-Dihydro-2H-pyrimido[1,2-c][1,3]benzothiazin-6-imine (PD 404,182, 1) is a known inhibitor
of bacterial 3-deoxy-D-manno-octulosonic acid 8-phosphate (KDO 8-P) synthase (Ki = 26 nM)
[1], Dimethylarginine dimethylaminohydrolase (DDAH1, IC50 = 9 µM) [2] and was also
described to be effective against hepatitis C virus (HCV, IC50 = 11 µM) and human
immunodeficiency virus (HIV, IC50 = 1 µM) [3], although the target and mechanism of anti-viral
action is not completely clear. Searching the Pubchem BioAssay Database revealed 7 other
proteins (histone lysine methyltransferase G9a, cytochrome P450 2C9, hydroxysteroid (17-beta)
dehydrogenase 4, 15-hydroxyprostaglandine dehydrogenase, human muscle pyruvate kinase,
aldehyde dehydrogenase 1, 15-human lipoxygenase 2) that were affected by 1 with bioactivity
values in the range from 0.11 to 14 µM [4]. 1 was inactive in 134 other bioassays. These
findings, suggest a certain potential for unselectivity of benzothiazin-6-imines. However, with
the exception of KDO 8-P synthase, the reported activities for mentioned proteins are primarily
in the micromolar range. Very recently, our group discovered 1 as an exceptional potent inhibitor
(IC50 = 11 nM) of human histone deacetylase 8 (HDAC8) and showed that it is highly selective
against the human isoenzymes HDAC1-7 [5]. The most potent analog shows an IC50-value of
2.9 nM against HDAC8. HDAC8 is an established target for T-cell lymphoma, childhood
neuroblastoma and related to several other types of tumor [6]. Our finding was completely
unexpected, since 1 lacks a classical zinc binding group that is a common feature of most known
inhibitors of zinc-dependent HDACs. Moreover, the structure-activity relationship of analog
compounds appeared to be very broad suggesting great potential for further optimization using
medicinal chemistry methods [5]. Okazaki et al. reported that the distinct S-N bond in the
isothiazolopyrimidine scaffold was immediately cleaved in the presence of GSH or other
4

ACCEPTED MANUSCRIPT
reducing agents giving the thiophenol 3, which was suggested to be the active anti-viral
component [7]. Ghebremariam et al. observed that 1 is a long persistent inhibitor of DDAH and
suggested that it may bind covalently to a cysteine residue in the catalytic site eventually forming
the mixed disulfide 6 [2]. We also observed a long lasting inhibitory effect of 1 on HDAC8:
About 90% of the enzyme remained inactive after 24 h dialysis. If the inhibitory mechanism
involved covalent modification, what would explain the noteworthy selectivity for HDAC8
against other targets and even closely related HDAC isoenzymes? And what could be learned
from the specific mode of action to develop optimized non-hydroxamate HDAC8 inhibitors?
These considerations motivated us to study the inhibition mechanism of HDAC8 by 1 in great
detail combining biochemical and biophysical methods with high-resolution HPLC-coupled ESI-
MS / MS techniques.
2. Materials and methods
Materials. All reagents and solvents were purchased from Sigma, Fluka, Bachem, Roth or
Tocris with a purity of 98% or higher. HPLC solvents were purchased from VWR with analytical
grade or better. Sequencing grade modified Trypsin lyophilized was purchased at Promega. To
minimize artificial oxidation of free thiols all measurements were performed in degassed buffer
solutions. To prevent oxidation during storage, enzymes were stored in the presence of 1 mM
TCEP. TCEP was removed immediately before measurements by GPC with Zeba Spin Desalting
Columns 7K MWCO (Thermo Scientific).
Mutagenesis. Mutant HDAC8 variants were generated using splicing by overlap extension PCR
(SOE-PCR) with the following Primers:
5

ACCEPTED MANUSCRIPT
HD8_BamHI_rev: 5’-AGGTGGATCCTTAAACAACGTGCTTCAGATTGCC-3’,
HD8_NdeI_for: 5’-GCGCATATGGAGGAGCCGGAGGAG-3’,
HD8_C102S_for: 5’-GGGCTAGGTTATGACTCCCCAGCCACTGAAGGGATA-3‘,
HD8_C102S_rev: 5‘-TATCCCTTCAGTGGCTGGGGAGTCATAACCTAGCCC-3‘,
HD8_C153S_for: 5‘-GATGAAGCATCTGGTTTTTCTTATCTCAATGATGCT-3‘,
HD8_C153S_rev: 5‘-AGCATCATTGAGATAAGAAAAACCAGATGCTTCATC-3‘.
DNA sequencing was performed at the sequencing service at the LMU Munich with the cycle,
clean and run (BigDye v3.1) protocol.
HDAC8 expression and purification. pET14b vector (Novagen, EMD Millipore) containing
codon-optimized human HDAC8, fused to a His6 SUMO tag, was used to express HDAC8 in E.
coli (BL21) DE3 pLysS. Cells were harvested by centrifugation for 10 min at 8000 g and 4 °C
and resuspended in lysis buffer (pH 8.0) containing 150 mM KCl, 50 mM Tris, 5 mM imidazole,
5 mM DTT, 1 x HALT protease inhibitor cocktail (Thermo Scientific) and 5 µg/mL DNAseI.
The cell suspension was sonicated and lysates were clarified by centrifugation at 18000 g for 30
min at 4 °C and sterile filtration. The filtrate was subsequently added to a 5 mL column volume
of cOmplete His tag purification resin (Roche), equilibrated with IMAC buffer A (pH 8.0)
containing 150 mM KCl, 50 mM Tris and 5 mM imidazole. After washing with 50 mL of the
same buffer His6-SUMO-HDAC8 was eluted with IMAC buffer B (pH 8.0) containing 150 mM
KCl, 50 mM Tris and 100 mM imidazole. Subsequently 10µg/ml His6 tagged SUMO-Protease
was added to the eluted HDAC8 fusion protein. Cleavage of His6-SUMO tag occurred overnight
whilst dialyzing against 25 mM Tris, 50 mM NaCl and 5 mM ß-ME (pH 8.0) at 4 °C. Then His6-
6

ACCEPTED MANUSCRIPT
SUMO tag and SUMO-Protease was removed by a second IMAC with AIC buffer A (pH 8.0)
containing 25 mM Tris and 50 mM NaCl and 5 mM imidazole. HDAC8 containing flowthrough
was concentrated and further purified on a strong anion exchanger (Bio-Scale Mini Macro-Prep
High Q 5 mL Cartridge, Biorad). After a washing step using AIC buffer A HDAC8 was eluted
using AIC buffer B (pH 8.0) containing 25 mM Tris and 1 M NaCl. 5 mM DTT was added to
prevent oxidation and remove possible ß-ME cysteine adducts. The final purification step
included size exclusion chromatog-raphy with a HiLoad Superdex 75 material (GE) equilibrated
with GPC Puffer (pH 8.0) containing 150 mM KCl and 50 mM Tris. The protein containing
fractions were collected and concentrated. Glycerol and TCEP were added to final
concentrations of 25 % and 1 mM and protein was stored at -20 °C. We typically obtained 3-5
mg HDAC8 from 1 liter culture.
Enzyme activity assay. All HDAC assays were executed in assay buffer (25 mM Tris-HCl pH
8.0, 50 mM NaCl and 0,001 % (v/v) pluronic F-68) in black half area 96-well microplates
(Greiner Bio-One, Germany). 1 nM of HDAC8WT or 10 nM of HDAC8 mutants were
preincubated with a serial dilution of the compounds for 1 h at 30 °C, before initiating the
enzyme reaction by addition of 10 µM Boc-Lys(TFA)-AMC. Some enzyme/compound mixtures
0
were treated with either 1 mM TCEP or various concentrations of GSH for 1h at 30 C, before
the enzyme reaction was started. The reaction was stopped after 60 min by adding 1.67 µM
SATFMK. The deacetylated substrate was cleaved with 0.42 mg/mL Trypsin to release
fluorescent AMC, which was detected with a microplate reader (PHERAstar FS, BMG
LABTECH) with fluorescence excitation at 360 nm and emission at 460 nm. IC50 values were
calculated by generating dose-response curves in GraphPad Prism 6 and fitting those to a 4-
parameter fit model. Sirtuin assays were performed according to a published procedue [8].
7

ACCEPTED MANUSCRIPT
Dialysis. HDAC8 (500 nM) was incubated with 100 µM 1 or buffer (control) for 1 h at room
temperature and dialyzed (MEMBRA-CEL® dialysis membrane MWCO 3500, Serva, Germany)
against 500-fold assay buffer with or without 50 mM ß-mercaptoethanol (-ME) for 22 h at 4 °C.
Samples were taken before and after dialysis to determine the activity of the protein, by adding
20 µM Boc-Lys(TFA)-AMC and incubating for 30 min at 30 °C. The reaction was stopped by
the addition of 1.7 µM SATFMK and 0,42 mg/mL Trypsin. Free AMC was detected in a
microplate reader with fluorescence excitation at 360 nm and emission at 460 nm.
Tryptic digestion. All tubes were cleaned with acetonitrile and water before use. 113 µM
HDAC8 was treated with 300 µM 1 for 1 h at 25 °C in 10 mM ammonium bicarbonate buffer
(ABC) while shaking at 400 rpm. A control sample was carried along without the addition of 1.
After treatment the samples were denatured at 90 °C for 30 min and alkylated with 14.5 mM
iodoacetamide for 1h at room temperature in the dark. All samples were diluted with ABC buffer
to 56.6 µM HDAC8 to accomplish a pH value between 7 and 8. Tryptic digestion was started
with 28.6 µg/mL sequencing grade modified trypsin (Promega, Germany) at 37 °C for 18 h while
shaking at 400 rpm. The digestion process was stopped by freezing and storing the samples at
-20 °C immediate after digestion till measurement.
Quantitative HPLC/ESI-QTOF-MS/MS. The liquid chromatography tandem mass
spectrometry (LC-MS/MS) system consisted of an Agilent 1100 high performance liquid
chromatograph (HPLC) equipped with a micro vacuum degasser, Multospher 120 RP 18-AQ,
250 x 4 mm (i.d.), 5 µm particle size coupled to an ABSciex X500R QTOF mass spectrometer
equipped with an IonDrive TurboV ESI-source. HPLC separation was performed under gradient
conditions with mobile phases of acetonitrile + 0.1% (v/v) formic acid (A) and Water + 0.1%
(v/v) formic acid (B) at a flow rate of 0.5 mL/min. The gradient condition was set as follows: 0-5
8

ACCEPTED MANUSCRIPT
min 10% A, 5-20 min linear increase to 70% A, 20-25 min 70% A, 25-26 min linear decrease to
the initial condition of 10% A, 26-35 min 10% A for column equilibration. The injection volume
was 10 µL. The parameters used for ESI-QTOF-MS/MS analysis in positive ion mode were as
follows: nebulizer (gas 1), 50 arbitrary units; turbo gas (gas 2), 55 arbitrary units; curtain gas, 35
arbitrary units; source temperature, 500°C; ionspray voltage + 5.5 kV, declustering potential
(DP), 50 V; collision energy (CE), 10 V. TOF MS full scan and product ion Data were acquired
by information dependent acquisition (IDA) scan mode in the mass range of 100-700 Da. The
calibration delivery system (CDS) was set to perform an automatic external calibration every ten
samples. Data acquisition and processing was performed with the Sciex OS 1.3 software.
The same system and mobile phases were used for peptide analysis but with different conditions
as mentioned before. A 100 x 2 mm (i.d.), 2.3 µm particle size TSKgel Super-ODS C18 column
with a flow rate of 0.2 mL/min was used with the gradient conditions set as follows: 0-5 min 5%
A, 5-40 min linear increase to 80% A, 40-45 min 80% A, 45-55 min linear decrease to 5 % A
and 55-60 min 5 % A for equilibration. The injection volume was 10 µL. ESI-QTOF-MS/MS
parameters were as follows: gas 1, 45 arbitrary units; gas 2, 45 arbitrary units; curtain gas, 35
arbitrary units; source temperature, 400°C; ionspray voltage + 5.5 kV, DP, 50 V; CE, 10 V. For
data acquisition IDA scan mode was used in the mass range of 450-2000 Da. Dynamic collision
energy mode was set active.
The software for Theoretical fragmentation and peptide fragment assignment were integrated in
Sciex OS 1.3.
Protein crystallography. Protein expression and crystallization of HDAH was performed as
described previously [9, 10]. Briefly, crystals of HDAH were grown by hanging drop vapor
9

ACCEPTED MANUSCRIPT
diffusion against 500 μl reservoir, the reservoir solution contained 0.1 M malate-imidazole
buffer (pH 5.25) and 3-6% PEG400. The reservoir solution and protein (10 mg/ml) were
mixed in a 1:1 ratio to a final volume of 4 μl. Crystals grew in three to ten days at 20°C. The
obtained crystals of HDAH were soaked by adding of 2µl of a 2 mM compound 1 dissolved in
the reservoir solution directly to the drop and incubated for 24h. Cryo-protection was achieved
by washing the crystal in 0.1 M malate-imidazole buffer with a pH of 5.25 and 30% PEG400.
The cryoprotected crystals were flash frozen in liquid nitrogen. Diffraction data were collected at
100 K on beamline PXII at the Swiss Light Source (SLS, Villigen, Switzerland).
Global fit analysis of reaction kinetics. The time courses of the relative concentrations of
molecular species 1-9 were obtained from quantitative ESI-MS-measurements as described
above. The simulation and modeling program COPASI was used to fit eligible models of the
reaction mechanism to the set of time resolved concentrations [11]. In short, the model is
described by a system of ordinary differential equations, which is determined by kinetic rate
constants. These parameters were optimized by starting with randomized start values for all
parameters and minimizing the total error sum of squares for the entire set of data curves by
subsequently application of the implemented Evolution Strategy, Evolutinary Programming and
Levenberg-Marquardt-algorithms. To estimate the absolute concentrations from relative
concentrations, proportional factors were concurrently fitted to the data. Entering the determined
rate constants and proportional factors into the model allowed for the simulation of time-courses
of all molecular species involved in the complex reaction mechanism.
Molecular docking. Modeling, preparation and visualization of structural data as well as
molecular docking was performed using MOE 2016 software (Chemical Computing Group ULC,
10

ACCEPTED MANUSCRIPT
Canada). At first crystal structure data PDB-ID 1T64, 1T69, 1VKG, 3SFF representing
significant conformations of HDAC8 in complex with various inhibitors were loaded into the
program and subjected to structure preparation including 3D protonation for subsequent docking.
The partial charges of all protein and ligand atoms were calculated using the implemented
Amber12 force field. Molecular docking was performed choosing the triangle matcher for
placement of the ligand in the binding site and ranked with the London dG scoring function. The
best 30 poses were passed to the refinement and energy minimization in the pocket using the
induced fit method and then rescored with the GBVI/WSA dG scoring function.
11

ACCEPTED MANUSCRIPT
3. Results and discussion
3.1. Compound 1 and closely related analogs show similar activity against HDAC8
1 been recently described as a novel, potent, and selective lead structure for HDAC8wt inhibitors
[5]. Most interestingly, the compound contains no common zinc-chelating group characteristic
for most HDAC inhibitors. Particularly the prevention of the ubiquitous hydroxamate group with
a high potential for unspecific zinc ion binding and suspected mutagenicity let the pharmacophor
of 1 appear an attractive alternative to existing HDAC8 inhibitors [12]. In light of its novel
structure highly dissimilar to known HDAC inhibitors, we wondered about the mode of action of
compound 1. Since 1 was reported to exhibit anti-viral properties, Okazaki et al. investigated the
chemical stability of the compound in greater detail and found out that 1 decomposes in the
presence of glutathione (GSH) into thiophenol 3 which was believed to be the active anti-viral
substance [7]. We confirmed the chemical instability of 1 in the presence of GSH, but besides 3
we also found considerable amounts of 5, 6, 8 and 9 using high-resolution ESI-MS. Next, we
noticed that thiophenol 3 decomposes into sulfenamide 5 and thiourethane 8 immediately after
dissolvation in 10 mM ammonium carbonate buffer (pH 8.0) (Fig. S1). To identify the active
substance that inhibits HDAC8wt, we determined IC50-values for analogs 1, 3 and 5, freshly
dissolved in assay buffer (Fig. 1A). Surprisingly, all three compounds turned out to be similarly
potent against HDAC8wt showing IC50-values of about 1 nM (Tab. S1). To simplify the
reporting of next results, we will talk about HDAC8 inhibition by 1 when this compound is
added to the enzyme, knowing that 1 can decompose into 3 and 5 in the presence of thiol groups
such as those in the side chain of solvent-accessible cysteines at the surface of HDAC8.
12

ACCEPTED MANUSCRIPT
3.2. Inhibition of HDAC8 is mediated by covalent modification of at least one cysteine
Since 1 inhibits in the concentration range of applied HDAC8wt, its binding is supposed to be
very tight. To distinguish between a covalent and a reversible binding mode, we performed a
dialysis of the complex formed by the interaction of HDAC8 and compound 1 (Fig. 2). After 22
hours of dialysis in the absence of reducing agent, the enzyme activity of HDAC8wt remains
substantially inhibited. In contrast, HDAC8 activity is completely recovered, if the protein-ligand
complex is dialyzed against buffer including -ME. This suggests covalent modification of
HDAC8wt by 1 involving the formation and fission of a disulfide bridge between at least one
accessible cysteine of the enzyme and the inhibitor. This conclusion is supported by the loss of
activity of compounds 1, 3 and 5 in the presence of the reducing agent TCEP that breaks
disulfide bonds (Fig. 1B, Tab. S1). Moreover, addition of TCEP to preformed complexes
between 1, 3 and 5 and HDAC8wt reverses enzyme inactivation confirming the reversibility of
the inhibitory mechanism that involves the formation and disruption of disulfide bonds. It should
be noted that GSH at concentrations up to 1 mM is not able to reverse the inactivation of
HDAC8wt once the disulfide with 1 is formed, possibly due to a less negative redox potential of
GSH (-230 mV compared with -290 mV of TCEP) (Tab. S1).
3.3. Mechanism of GSH-modification by 1
The reaction of GSH and 1 was investigated previously to learn about the corresponding
transformations in aqueous solution. Okazaki et al. found that 1 was immediately cleaved in the
presence of GSH producing 3 [7]. Similar to our results, 1 and 3 showed comparable EC50-values
for the inhibition of HIV-1 infection. From these results, 1 was suggested as prodrug for the
13

ACCEPTED MANUSCRIPT
bioactive thiophenol 3 [7]. We applied the powerful technique of high-resolution ESI-MS to
prove the existence of nine occurring molecular species in the reaction of GSH and 1 (Fig.’s 3,
S2, S3, S5 and S6). In contrast to Okazaki, we verified only very small amounts of 3
immediately after mixing 1 and GSH. Similarly, purified 3 also disappears instantly after
dissolvation in ammonium carbonate buffer pH 8.0 generating significant amounts of
sulfenamide 5 and only minor amounts of 9 (Fig. S1). Therefore, we conclude that 3 is
chemically unstable in aqueous solution. The concentration of every proven molecular entities
involved in the reaction between GSH and 1 was quantified using LC-ESI-MS, and the
corresponding time courses analyzed during a period of 15 h to elucidate the accurate mechanism
of this unexpected complex reaction (Fig. 4 A, B). The decay of 1 shows two phases: a rapid
drop within the first hour of reaction and a slow decrease over hours. In contrast, GSH
disappears completely from the mixture within 30 minutes. Two intermediates, 5 and 6 are
generated rather rapidly reaching their peak concentration at about 2 h and 0.5 h, respectively.
While the mixed disulfide 6 vanishes after 3 h, the sulfenamide 5 is unexpected metastable and
still present as one of the dominating species after 15 h. GSSG appears in conjunction with the
rapid drop in GSH concentration and subsequently decreases slowly approaching a rather stable
level. The occurrence of glutathione-thiocyanate (4) was unequivocally proven by ESI-MS/MS
data (Fig. S2,S3), removal of the cyano group and the characteristic formation of a cyclic
thiazolamine after reacting with a 5-fold excess concentration of cysteine (Fig. S4). Interestingly,
4 was shown to arise rapidly and be chemically stable for at least 15 h (Fig. 4A). Simulations
suggest that cyanylation is the predominant modification of GSH compared with disulfide
formation (Fig. 4B). The whole complex reaction is superimposed by slowly emerging 8 and
14

ACCEPTED MANUSCRIPT
sulfinic acid 9 as final products. The array of time-dependent concentrations of all species was fit
to a model taking all experimentally verified molecular entities into account (Fig. 4C).
The entire set of experimental time-dependent concentrations is well described by the reaction
scheme in Fig. 4C). Comprehensive data analysis using COPASI provides the rate constants for
every depicted elementary reaction steps (Tab. S2). The reaction mechanism postulates the
formation of glutathione thiocyanate, 4, proven be ESI-MS/MS, as suggested by Ghebremariam
for the covalent inactivation of DDAH1 [2]. In the same step thiophenol 3 is generated that is
instantly oxidized (see above) giving the central metastable sulfenamide 5. The mixed disulfide 6
generated from 5 can react further with GSH giving GSSG and 3. When GSH is consumed, the
chemical equilibrium between 5, 6 and 7 is established leading to a progressive drop in the
concentration of 6 that is no longer produced from 5 and a slower moderate decrease of GSSG
concentration. Simultaneously, two slow side reactions take place generating two chemically
stable final products: Metastable 5 is oxidized to sulfinic acid 9, and residual 1 is hydrolyzed to
thiourethane 8. The comprehensive analysis of quantitative and time-resolved data provide
substantial insight into the accurate mechanism by which model compound GSH and 1 react.
From these results we conclude that 1 is a general dual cysteine modifier in aqueous solution that
can transfer cyano groups to thiol residues and also produces mixed disulfides between cysteines
and thiophenol 3.
3.4. Covalent modification of HDAC8wt by 1
In an attempt to elucidate the accurate binding mode of 1 to HDAC enzymes we tried to
crystallize complexes of 1 with HDAC8 and HDAH, a bacterial homolog that crystallizes nicely
in our hands [9, 10]. We added 1 to a HDAH crystal in order to soak the ligand into the binding
pocket. Unfortunately, we were not successful to obtain crystals of the HDAC8 complex with
15

ACCEPTED MANUSCRIPT
sufficient quality for structure determination. But we were able to determine the structure of the
reaction product of HDAH and 1 at a very high spatial resolution of 1.47 Å. To our surprise, the
binding site pocket was unoccupied. Instead, we found covalent modifications of two cysteines
(C51 and C295) at the surface of HDAH.
Although no reducing agent was present during crystallization, both modifications were clearly
visible as mixed disulfides between the respective cysteine and thiophenol 3 as shown exemplary
for C51 (Fig. 5). The observed modifications of HDAH are in agreement with the reaction
mechanism between 1 and GSH (Fig. 4C), which has been elucidated on the basis of quantitative
ESI-MS data. The cysteine modification reaction seems to be rather unspecific and is supposed
to be a common mechanism for unselective inhibition of other enzymes, if functional cysteines
are hit.
To identify putatively modified cysteines in HDAC8wt, a reaction mixture of HDAC8wt and 1
was subjected to tryptic digestion followed by LC-ESI-MS/MS analysis. In fact, C153 in the
catalytic site as well as various other cysteine residues were modified by 1 (Tab. S3). Five
cysteines, including C28, C153, C244, C314 and C352 showed cyanylation upon treatment with
1 (Fig. S3). These modifications are in agreement with the unspecific affinity labeling
mechanism of GSH. Since the reaction of HDAC8wt with 1 was performed at high concentrations
of almost four orders of magnitude above the IC50-value of 1, only qualitative statements about
the modified cysteines in HDAC8wt and no assessment of a preference for a particular cysteine
can be made. However, it appears plausible that the effect of 1 or 5 responsible for the inhibition
of HDAC8wt is mainly due to the modification of C153 in the catalytic site pocket thereby
permanently blocking access to the catalytic center. In any case, 1 and most likely also related
benzothiazine-imines are clearly able to modify solvent accessible thiol groups on protein
16

ACCEPTED MANUSCRIPT
surfaces and produce cysteine-thiocyanates as well as mixed disulfides between cysteine and 3.
For the majority of enzyme targets including other HDAC isoenzymes that are effected by 1 only
moderate activities in the micromolar range have been reported [2-5]. 1, 3 and 5 were also tested
against Sirtuins, because Sirt1 was shown to respond to similar inhibitors [13]. The compounds
show moderate micromolar activities against Sirt1 and Sirt2, but were virtually inactive against
Sirt 3 (Tab. S7). This is in agreement with moderate activity on a variety of protein targets and
our finding that 1 is a general cystein modifier. Most recently, thiol reactive histone
acetyltransferase (HAT) inhibitors have been considered as essentially nonselective interference
compounds [13]. Typically, these HAT inhibitors show moderate IC50-values in the micromolar
range. However, CPTH2, a thiol-reactive HAT (GCN5) inhibitor, does not inhibit HDAC8 (IC50
> 50 µM, Fig. S13). Despite its unspecific cysteine reactivity, 1 is highly potent and selective
against HDAC8 with an IC50-value of 11 nM [5]. Since C153 is conserved among all human
HDAC isoenzymes and other cysteins in HDAC8 are also modified, another component had to
be the major contributor to the observed selectivity of compound 1.
3.5. C102 and C153 are involved in the inhibition mechanism of HDAC8
Looking at available crystal structures of HDAC8 (e.g. PDB-ID: 1T69) reveals C153 in the
active site and C102 in close proximity to be eligible for a potential chemical modification by 1.
In the following, single mutants HDAC8C102S and HDAC8C153S and the double mutant
HDAC8C102S/C153S were produced to elucidate the participation of the exchanged cysteines in the
mechanism of inhibition. Since all HDAC8 variants were enzymatically active, the inhibitory
efficacy of compounds 1, 3 and 5 could be determined in terms of IC50-values (Fig. 1A, Fig.
S12). To demonstrate the reversibility of a putative disulfide bond involving C102 or C153,
17

ACCEPTED MANUSCRIPT
respectively, the concentration series were performed in the absence or after the addition of 1
mM TCEP to the preformed complex between HDAC8-variant and compound (Fig. S12). After
the addition of the reducing agent, the enzyme activity of all HDAC8 variants is recovered
confirming that the mechanism of HDAC8 inhibition by 1 and derivatives thereof involves the
formation of disulfide bonds (Fig. S12). SAHA was used as control that shows absolutely no
susceptibility to reducing agent TCEP. IC50-values of SAHA varied slightly between 0.29 and
2.4 µM among the 4 HDAC8 variants (Tab. S1). While SAHA behaves like a classical active site
inhibitor against the double mutant HDAC8C102S/C153S, compounds 1, 3 and 5 show a biphasic
dose-response curve against the same enzyme (Fig. S12G). This finding suggests an allosteric
interaction site at this HDAC8 variant. The IC50-values of 1, 3 and 5 for the more potent site is
comparable with the IC50-values against the single mutants of HDAC8. We hypothesize that
other solvent accessible cysteines at the surface of HDAC8 may react with analogs of 1 enabling
an indirect modulation of the catalytic efficiency of HDAC8C102S/C153S, which is not capable of
full enzyme inactivation. Only at very high concentration of 1, 3 or 5 (> 10 µM) full inhibition is
achieved probably by additional unselective modifications of the HDAC8 variant. These results
suggest that HDAC8 contains at least two types of interaction sites: A high affinity site involving
cysteins C102 and C153 in close proximity to the active site and at least one low affinity
allosteric site where unselective disulfide modification can take place at higher concentrations of
1 analogs. There are only rather small differences in IC50 values of 1, 3 and 5 between HDAC8wt
and HDAC8C153S and HDAC8C102S (Tab. S1). Due to the low enzyme activity of the HDAC8
mutant proteins as much as 10 nM enzyme concentration was chosen to measure and compare
the dose-response curves of the inhibitors against all HDAC8 variants. Since particularly the
IC50-values with HDAC8wt approach the level of enzyme concentration, these IC50-values only
18

ACCEPTED MANUSCRIPT
represent upper limits of the true inhibitory activity. This is also recognizable by the distinct
steep shape of the dose-response curves of 1, 3, 5 on HDAC8wt indicating nearly stoichiometric
tight binding of a ligand (Fig. S12 A). Nevertheless, a clear differentiation between the inhibition
of HDAC8wt and the mutant enzymes is not possible on the basis of these data. Unfortunately, it
is not possible to lower the enzyme concentration of the mutant variants because of their weak
enzyme activity. Moreover, it is more conclusive to investigate the inhibition kinetics of covalent
inactivators such as cystine modifiers. In fact, the inhibition kinetics of HDAC8wt by 1 is clearly
different from that of the mutant variants (Fig. 6). HDAC8wt is essentially completely inhibited
after 2 minutes in striking contrast to HDAC8C102S, HDAC8C153S and HDAC8C102S/C153S
indicating the pivotal role of both cysteins in the inhibition mechanism. 1 shows a slow
inhibition of the double mutant HDAC8 approaching an equilibrium level of about 50% residual
enzyme activity, which is in agreement with the corresponding dose-response curve (Fig. S12G).
This finding suggests that the inhibition mechanism does not involve a continuing chemical
inactivation but is rather maintained by non-covalent binding. In contrast, inhibition of
HDAC8C153S and HDAC8C102S does not reach equilibrium but still continues even after 30 min of
reaction suggesting a covalent modification of C102 and C153 (Fig. 6). Interestingly, C102
seems to be more rapidly hit by 1 than C153. The enzyme activity of HDAC8C153S drops
instantly by about 50% and than shows a slow ongoing inhibition, whereas HDAC8C102S shows
only the slow inhibition kinetics. The differential inhibition kinetics of the HDAC8 variants
clearly demonstrates, that the presence of both cysteins, C102 and C153, is required for the rapid
inhibition mechanism of HDAC8wt.
3.6. Non-covalent molecular recognition of 1 and 5 by HDAC8wt
19

ACCEPTED MANUSCRIPT
The reactivity of 1 in the presence of GSH and the unselective modification of many proteins
render benzothiazine-imines unsuitable for therapeutic applications. On the other hand, the
significant increase in potency of 1 against HDAC8 by three orders of magnitudes cannot be
explained by simple nonspecific affinity labeling. Therefore, we hypothesize that the
experimentally observed high selectivity of benzothiazine-imine analogs for HDAC8 is caused
by an additional specific contribution to inhibition such as a beneficial non-covalent molecular
recognition in the active site or another transient binding pocket that precedes covalent
modification. Our hypothesis is supported by the inhibition kinetics of the HDAC8C102S/C153S
double mutant that is in agreement with non-covalent binding (Fig. 6). The presumed non-
covalent molecular recognition of 1 and 5 by HDAC8 was investigated further using
computational methods, particularly molecular docking. It is known that the conserved binding
pocket of HDAC8 is highly conserved among members of the HDAC family. However, HDAC8
is very special with respect to enhanced malleability of the region around the classical active site.
There exist various X-ray structures that demonstrate the considerable flexibility of HDAC8 and
a transition from closed (PDB-ID: 1T69) to sub-open (PDB-ID: 1T64) to wide-open (1VKG)
binding pockets [14]. Most remarkable, the sub-open state shows a deep second binding pocket
next to the conserved one, which is occupied with a second inhibitor molecule. The L1-and L2-
loop are primarily responsible for ligand-induced conformational changes in HDAC8 [14-16].
Both loops display highly varying amino acid compositions in HDAC1 and HDAC8. In
addition, the L1-loop of HDAC1 is 5 amino acids longer than that of HDAC8 causing a different
flexibility profile in both isoenzymes [17]. Compounds 1 and 5 were docked into the binding
pockets of above mentioned representative HDAC8 structures to determine the favored binding
modes and identify putative beneficial contacts. Furthermore, 1 was also docked into available
20

ACCEPTED MANUSCRIPT
crystal structures of HDAC isoenzymes HDACs 1, 2, 4 and 6 to understand the observed
selectivity in enzyme activity assays on a molecular basis (Fig. S9). The best docking score is
obtained for HDAC8, thus confirming the experimental data (Tab. S4).
Docking of 1 into various crystal structures of HDAC8 revealed clearly different scores
indicating different affinities to distinct conformations of HDAC8 (Tab. S5). The best docking
pose of 1 at the bottom of the active site pocket in HDAC8 (PDB-ID: 1T69) is shown in Fig. 7A.
A comparison of the docking result between HDAC8 (PDB-ID: 1T69) and 1 with the crystal
structure of HDAC8 bound to an amino-acid inhibitors (PDB-ID: 3SFF) shows a substantial
overlap of both inhibitors binding to the transition area between the acetyl-lysine binding tunnel
and the acetate release channel (Fig. 7B). The best docking pose of 1 in HDAC1 (PDB-ID:
5ICN) is located in the conserved active site binding pocket (Fig. S9). The closest distance
between carbon atoms of F150 (matching F152 in HDAC8) and Y303 (matching Y306 in
HDAC8) is 4.0 Å impeding access to the acetate release channel (Fig. S10). In addition, M30
located in the L1-loop of HDAC1 intrudes into the acetate release channel and clashes with
superimposed compound 1 docked to HDAC8. In contrast, F152 and Y306 in the docked
complex between HDAC8 and 1 are moved to the side allowing the aromatic moiety of 1 to
protrude into the acetate release channel similar to the amino-acid derived inhibitors of
Whitehead et al. [18] (Fig. 7 and S10). The dominating molecular interactions between HDAC8
and 1 are zinc binding through the imine nitrogen of 1 and a hydrogen bond between a nitrogen
atom of the tetrahydropyrimidine ring of 1 and Y306. These are complemented by a cation--
interaction between the aromatic ring of 1 and R37 and mainly hydrophobic interactions to F152,
F208, C153, I34 and G304. The aromatic ring of 1 forms a π-π T-shaped interaction with W141
(Fig. S10). Acetate release is believed to be mediated by R37 stabilizing the charge of acetate in
21

ACCEPTED MANUSCRIPT
cooperation with W141 relocating its indole moiety to close the release channel [18, 19]. The
pivotal W141 is substituted by leucine in class I HDACs 1, 2 and 3, and has no counterpart in
class II HDACs. This amino acid difference in the acetate release channel and also in the L1-
loop (e.g. M30 in HDAC1 and 2 intrudes into the acetate release channel) as well as in the L2-
loop are supposed to have significant impact on HDAC8 flexibility and the geometry of the
internal binding cavity. In summary, differences in pivotal amino acids, protein malleability and
the particular binding mode of 1 to HDAC8 is suggested to be the reason for the observed
selectivity for HDAC8 and against class I HDACs 1 and 2 as well as class IIa HDAC4 and class
IIb HDAC6. Once 1 is non-covalently bound to HDAC8, it is brought in close proximity to C153
lining the active site pocket (Fig. S10) and is supposed to have an increased probability to
cyanylate this cysteine (Fig. 8). The released thiophenol 3 is rapidly oxidized to 5 in free
solution. But in principle 3 could also remain transiently bound and stabilized by non-covalent
interactions before reacting further with the adjacent C153-thiocyanate (Fig. 8). The sulfur atom
of the thiocyanate possesses an electrophilic character due to the cyano moiety acting as a
leaving group [20]. The experimentally observed mixed disulfide can be reached by substituting
the cyanyl group by thiophenol 3.
Docking 3 into HDAC8 with cyanylated C153 shows that 3 is well accommodated in the binding
pocket and has a short distance of 4.7 Å between the thiol group and the sulfur atom of C153
(Fig. S7). During the first hour of interaction between 1 and GSH the next prominent reactive
intermediated during the reaction with thiol agents is the sulfenamide 5 which is rapidly formed
by oxidation of 3 and rather stable under assay conditions. In fact, 5 showed similar potency
against HDAC8 than 1. Therefore, non-covalent binding of 5 followed by direct conversion of
C153 to the mixed disulfide with 3 could be an alternative pathway to inhibition of HDAC8
22

ACCEPTED MANUSCRIPT
according to the model reaction between GSH and 1 (Fig. 8). Docking of 5 to HDAC8 revealed a
binding mode, where a partially negatively charged nitrogen of 5 binds closely to the zinc ion
(Fig. S8). The non-covalent interaction is enhanced by a hydrogen bond to Y306 and various
hydrophobic interactions to C153, W141, H142, H143, F208, M274, Y306 and G304. The thiol
group of C153 and the sulfur atom of 5 are brought in proximity (4.8 Å) allowing for a
subsequent modification reaction (Fig. S8, Tab. S4). Since ESI-MS/MS data prove the
cyanylation of C153, we conclude that the interaction between 1 and HDAC8 follows the
cyanylation pathway (Fig.8 upper panel). However, since 1 is not so stable in aqueous solution
and 5 is also capable to inhibit HDAC8 with comparable potency, we cannot rule out the second
pathway which starts with the interaction between metastable 5 and HDAC8 producing a mixed
disulfide that blocks access to the active site (Fig. 8 lower panel).
23

ACCEPTED MANUSCRIPT
4. Conclusions
The benzothiazine-imine 1 is a potent and selective inhibitor of HDAC8 with an unprecedented
chemical scaffold lacking traditional zinc binding groups. This study provides detailed insight in
the reaction mechanism of 1 with GSH on the basis of quantitative and time-resolved ESI-
MS/MS data. Compound 1 is chemically not stable in the presence of GSH. The reaction
produces thiocyanate and disulfide modifications of GSH, GSSG as well as a mixed disulfide
with 3. Surprisingly, the sulfenamide 5 appears to be relatively stable over the course of 15 h at
pH 8.0.
Combined MS, biochemical and crystal structure data of the interaction of HDAC8 with 1
provide clear evidence that 1) 1 is a covalent inhibitor of HDAC8, 2) inhibition is reversible in
the presence of reducing agents, 3) C153 in the active site and adjacent C102 are involved in the
inhibition mechanism and 4) 1 modifies various cysteines in HDAC8 forming either thiocyanates
or mixed disulfides with 3.
The reactive nature of 1 is obviously the cause for its moderate inhibitory effects against several
other targets. However, the single-digit nanomolar potency and high selectivity of 1 for HDAC8
even against closely related HDAC isoenzymes requires a further important component in the
specific recognition of 1 by HDAC8.
On the basis of molecular docking results we propose putative selective non-covalent binding of
1 and 5 to HDAC8 prior to covalent modification. The optimal binding mode of 1 at the bottom
of the active site protrudes into the acetate release channel showing distinct amino acid
differences to other HDAC isoenzymes. The close proximity of 1 and C153 within the active site
pocket is supposed to accelerate the covalent inactivation particularly in HDAC8 and is proposed
to be a major determinant for the observed selectivity.
24

ACCEPTED MANUSCRIPT
It is quite clear, that the reactive nature of 1 and its analogs render them unsuitable for
therapeutic treatment. However, it appears attractive to further examine the specific molecular
recognition motif of 1 that utilizes the transition area between the catalytic site and acetate
release channel. We are currently pursuing several strategies towards less reactive variants of 1
that exploit the specific non-covalent molecular recognition of this class of compounds by
HDAC8.
25

ACCEPTED MANUSCRIPT
Conflict of interest
The authors declare no competing financial interest.
Acknowledgement
NW and MJ thank the Deutsche Forschungsgemeinschaft (DFG,
Ju295/14-1) for funding.
Appendix A. Supplementary data
Figure Captions:
Fig. 1: Dose-response curve of 1, 3, 5 and control SAHA against indicated HDAC8wt. 1 nM
HDAC8wt was A) in the absence and B) in the presence of TCEP incubated with varying
concentrations of indicated inhibitor for 1 h at 30 °C. The remaining HDAC8wt-acitivity was
determined using the standard enzyme activity assay. All data points represent means and
standard deviations from 3 independent experiments. Solid lines are best-fit to a 4-parameter
logistic.
Fig. 2: Irreversible binding of 1 to HDAC8wt becomes reversible upon addition of reducing agent
-mercapto ethanol (-ME). Enzyme activity is referred to 100 % free HDAC8wt. After addition
of 100 µM 1 to 500 nM HDAC8wt the enzyme is completely inhibited (before dialysis). After 24
h dialysis at 40C in the absence of -ME only weak and in the presence of 50 mM -ME almost
the entire enzyme activity is recovered. Shown data are means and standard deviations of three
independent experiments.
Fig. 3: Quantitative ESI-MS of all observed educts, intermediates and products of the reaction
between compound 1 and GSH (2). The bold numbers under the m/z ratios correspond to those
in Fig. 4 of the main text. The spectra are measured after indicated time periods.
Fig. 4: A) Time courses of indicated species during the reaction between 2.3 µM 1 and 1.6 µM
GSH in 10 mM ammonium carbonate buffer pH 8.0 at (26 ± 1) °C. Experimental relative
concentrations as determined from ESI-MS. Solid lines represent a simultaneous fit of the
reaction mechanism shown in C) to the data using COPASI. B) Simulated absolute
concentrations on the basis of the reaction mechanism and rate constants determined in C) using
COPASI. C) Proposed reaction mechanism of 1 with glutathione (GSH). GSSG is oxidized GSH
26

ACCEPTED MANUSCRIPT
and 4 is glutathione-thiocyanate. All chemical species have been identified and quantified using
ESI-MS (Fig. S2).
Fig. 5: Covalent modification of C51 in a crystal structure of HDAH (PDB-ID: 6GJK) after
reaction with 1. C51 forms a mixed disulfide bridge with thiophenol 3. The blue mesh shows the
2Fo-Fc-electron density map of the compound contoured at 1 sigma.
Fig. 6: Inhibition kinetics of different HDAC8 variants by compound 1. 100 nM of the respective
HDAC8 variant was incubated with 500 nM of freshly dissolved inhibitor at 300C. The residual
enzyme activity was measured after varying incubation time periods. The data represent mean
and standard deviation of at least four independent experiments.
Fig. 7: Docking of 1 into the active site pocket of HDAC8 (PDB-ID: 1T69). A) 1 binds at the
bottom of the active site pocket and protrudes into the acetate release channel. The green dashed
line indicates a hydrogen bond to Y306 and the magenta dashed line a metal-binding interaction
between the imine group and the catalytic zinc ion. B) Overlay of 1 (orange) docked into PDB-
ID 1T69 and the complex between HDAC8 and amino-acid based inhibitor (PDB-ID: 3SFF,
green) [18]. Both compounds show pronounced overlap, particularly in the transition area
between the acetyl-lysine binding tunnel and the horizontal acetate release channel.
Fig. 8: Two pathways of covalent modification at C153 in the active site of HDAC8 by 1 or 5
resulting in thiocyanate or mixed disulfide, respectively.
27

ACCEPTED MANUSCRIPT
References
[1] M.R. Birck, T.P. Holler, R.W. Woodard, Identification of a Slow Tight-Binding Inhibitor of
3-Deoxy-d-m anno-octulosonic Acid 8-Phosphate Synthase, J. Am. Chem. Soc., 122 (2000)
9334-9335.
[2] Y.T. Ghebremariam, D.A. Erlanson, J.P. Cooke, A novel and potent inhibitor of
dimethylarginine dimethylaminohydrolase: a modulator of cardiovascular nitric oxide, J.
Pharmacol. Exp. Ther., 348 (2014) 69-76.
[3] A.M. Chamoun, K. Chockalingam, M. Bobardt, R. Simeon, J. Chang, P. Gallay, Z. Chen, PD
404,182 is a virocidal small molecule that disrupts hepatitis C virus and human
immunodeficiency virus, Antimicrob. Agents Chemother., 56 (2012) 672-681.
[4] National Center for Biotechnology Information. PubChem Compound Database;
CID=6603967, https://pubchem.ncbi.nlm.nih.gov/compound/6603967 (accessed Apr. 21, 2018).
in.
[5] A. Kleinschek, C. Meyners, E. Digiorgio, C. Brancolini, F.J. Meyer-Almes, Potent and
Selective Non-hydroxamate Histone Deacetylase 8 Inhibitors, ChemMedChem, 11 (2016) 2598-
2606.
[6] A. Chakrabarti, J. Melesina, F.R. Kolbinger, I. Oehme, J. Senger, O. Witt, W. Sippl, M. Jung,
Targeting histone deacetylase 8 as a therapeutic approach to cancer and neurodegenerative
diseases, Future medicinal chemistry, 8 (2016) 1609-1634.
[7] S. Okazaki, S. Oishi, T. Mizuhara, K. Shimura, H. Murayama, H. Ohno, M. Matsuoka, N.
Fujii, Investigations of possible prodrug structures for 2-(2-mercaptophenyl)
tetrahydropyrimidines: reductive conversion from anti-HIV agents with pyrimidobenzothiazine
and isothiazolopyrimidine scaffolds, Org. Biomol. Chem., 13 (2015) 4706-4713.
[8] B. Heltweg, J. Trapp, M. Jung, In vitro assays for the determination of histone deacetylase
activity, Methods, 36 (2005) 332-337.
[9] C. Meyners, A. Kramer, O. Yildiz, F.J. Meyer-Almes, The thermodynamic signature of
ligand binding to histone deacetylase-like amidohydrolases is most sensitive to the flexibility in
the L2-loop lining the active site pocket, Biochim. Biophys. Acta, 1861 (2017) 1855-1863.
[10] C.E. Weston, A. Kramer, F. Colin, O. Yildiz, M.G. Baud, F.J. Meyer-Almes, M.J. Fuchter,
Toward Photopharmacological Antimicrobial Chemotherapy Using Photoswitchable
Amidohydrolase Inhibitors, ACS Infect Dis, 3 (2017) 152-161.
[11] S. Hoops, S. Sahle, R. Gauges, C. Lee, J. Pahle, N. Simus, M. Singhal, L. Xu, P. Mendes, U.
Kummer, COPASI--a COmplex PAthway SImulator, Bioinformatics, 22 (2006) 3067-3074.
[12] S. Shen, A.P. Kozikowski, Why Hydroxamates May Not Be the Best Histone Deacetylase
Inhibitors-What Some May Have Forgotten or Would Rather Forget?, ChemMedChem, 11
(2016) 15-21.
[13] J.L. Dahlin, K.M. Nelson, J.M. Strasser, D. Barsyte-Lovejoy, M.M. Szewczyk, S. Organ, M.
Cuellar, G. Singh, J.H. Shrimp, N. Nguyen, J.L. Meier, C.H. Arrowsmith, P.J. Brown, J.B. Baell,
M.A. Walters, Assay interference and off-target liabilities of reported histone acetyltransferase
inhibitors, Nature Communications, 8 (2017) 1527.
[14] J.R. Somoza, R.J. Skene, B.A. Katz, C. Mol, J.D. Ho, A.J. Jennings, C. Luong, A. Arvai,
J.J. Buggy, E. Chi, J. Tang, B.C. Sang, E. Verner, R. Wynands, E.M. Leahy, D.R. Dougan, G.
Snell, M. Navre, M.W. Knuth, R.V. Swanson, D.E. McRee, L.W. Tari, Structural snapshots of
human HDAC8 provide insights into the class I histone deacetylases, Structure, 12 (2004) 1325-
1334.
28

ACCEPTED MANUSCRIPT
[15] M. Brunsteiner, P.A. Petukhov, Insights from comprehensive multiple receptor docking to
HDAC8, J. Mol. Model., 18 (2012) 3927-3939.
[16] D.P. Dowling, S.L. Gantt, S.G. Gattis, C.A. Fierke, D.W. Christianson, Structural Studies of
Human Histone Deacetylase 8 and Its Site-Specific Variants Complexed with Substrate and
Inhibitors, Biochemistry, 47 (2008) 13554‐13563.
[17] N. Deschamps, C.A. Simões-Pires, P.-A. Carrupt, A. Nurisso, How the flexibility of human
histone deacetylases influences ligand binding: an overview, Drug Discov. Today, 20 (2015)
736-742.
[18] L. Whitehead, M.R. Dobler, B. Radetich, Y. Zhu, P.W. Atadja, T. Claiborne, J.E. Grob, A.
McRiner, M.R. Pancost, A. Patnaik, W. Shao, M. Shultz, R. Tichkule, R.A. Tommasi, B. Vash,
P. Wang, T. Stams, Human HDAC isoform selectivity achieved via exploitation of the acetate
release channel with structurally unique small molecule inhibitors, Bioorg. Med. Chem., 19
(2011) 4626-4634.
[19] S. Haider, C.G. Joseph, S. Neidle, C.A. Fierke, M.J. Fuchter, On the function of the internal
cavity of histone deacetylase protein 8: R37 is a crucial residue for catalysis, Bioorg. Med.
Chem. Lett., 21 (2011) 2129-2132.
[20] T. Castanheiro, J. Suffert, M. Donnard, M. Gulea, Recent advances in the chemistry of
organic thiocyanates, Chemical Society Reviews, 45 (2016) 494-505.
29