Design, Synthesis, Molecular Modeling, and Biological Evaluation of Novel Amine-based Histone Deacetylase Inhibitors

Article abstract of DOI:10.1002/cmdc.201700449

Histone deacetylases (HDACs) are promising drug targets for a variety of therapeutic applications. Herein we describe the design, synthesis, biological evaluation in cellular models of cancer, and preliminary drug metabolism and pharmacokinetic studies (DMPK) of a series of secondary and tertiary N-substituted 7-aminoheptanohydroxamic acid-based HDAC inhibitors. Introduction of an amino group with one or two surface binding groups (SBGs) yielded a successful strategy to develop novel and potent HDAC inhibitors. The secondary amines were found to be generally more potent than the corresponding tertiary amines. Docking studies suggested that the SBGs of tertiary amines cannot be favorably accommodated at the gorge region of the binding site. The secondary amines with naphthalen-2-ylmethyl, 5-phenylthiophen-2-ylmethyl, and 1H-indol-2-ylmethyl (2 j) substituents exhibited the highest potency against class I HDACs: HDAC1 IC₅₀ 39–61 nm, HDAC2 IC₅₀ 260–690 nm, HDAC3 IC₅₀ 25–68 nm, and HDAC8 IC₅₀ 320–620 nm. The cytotoxicity of a representative set of secondary and tertiary N-substituted 7-aminoheptanoic acid hydroxyamide-based inhibitors against HT-29, SH-SY5Y, and MCF-7 cancer cells correlated with their inhibition of HDAC1, 2, and 3 and was found to be similar to or better than that of suberoylanilide hydroxamic acid (SAHA). Compounds in this series increased the acetylation of histones H3 and H4 in a time-dependent manner. DMPK studies indicated that secondary amine 2 j is metabolically stable and has plasma and brain concentrations >23- and >1.6-fold higher than the IC₅₀ value for class I HDACs, respectively. Overall, the secondary and tertiary N-substituted 7-aminoheptanoic acid hydroxyamide-based inhibitors exhibit excellent lead- and drug-like properties and therapeutic capacity for cancer applications.

Full text of DOI:10.1002/cmdc.201700449

Accepted Article
Title: Design, Synthesis, Molecular Modeling, and Biological Evaluation
of Novel Amine-based Histone Deacetylase Inhibitors
Authors: Pavel A. Petukhov, Hazem Abdelkarim, Raghupathi
Neelarapu, Antonett Madriaga, Irida Kastrati, Yue-ting Wang,
Aditya S. Vaidya, Taha Y. Taha, Gregory R. J. Thatcher, and
Jonna Frasor
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemMedChem 10.1002/cmdc.201700449
Link to VoR: http://dx.doi.org/10.1002/cmdc.201700449
A Journal of
www.chemmedchem.org

10.1002/cmdc.201700449
ChemMedChem
Design, Synthesis, Molecular Modeling, and Biological
Evaluation of Novel Amine-based Histone Deacetylase
Inhibitors
Dr. Hazem Abdelkarim,† Dr. Raghupathi Neelarapu,† Antonett Madriaga,† Dr. Aditya S. Vaidya,†
Dr. Irida Kastrati,‡ Dr. Yue-ting Wang,† Taha Y. Taha,† Prof. Gregory R. J. Thatcher,† Prof. Jonna
Frasor,‡ Prof. Pavel A. Petukhov†,*
†
Department of Medicinal Chemistry and Pharmacognosy, College of Pharmacy, University of
Illinois at Chicago, 833 South Wood Street, Chicago, IL 60612, USA.
‡ Department of Physiology and Biophysics, University of Illinois at Chicago, Chicago, IL 60612,
USA.
* Corresponding author: Phone: 312-996-4174. Fax: 312-996-7107. E-mail: pap4@uic.edu
1
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
ABSTRACT: Histone deacetylases (HDACs) are promising drug targets for a variety of
therapeutic applications. Here we describe the design, synthesis, biological evaluation in cellular
models of cancer, and preliminary drug metabolism and pharmacokinetic studies (DMPK) of a
series of secondary and tertiary N-substituted 7-aminoheptanoic acid hydroxyamide-based HDAC
inhibitors 2 and 3, respectively. Introduction of an amino group with one or two surface binding
groups (SBGs) yielded a successful strategy to develop novel and potent HDAC inhibitors.
Secondary amines 2 were found to be generally more potent than the corresponding tertiary amines
3. Docking studies suggested that the SBGs of tertiary amines 3 cannot be favorably
accommodated at the gorge region of the binding site. The secondary amines with naphthalen-2-
ylmethyl (2g), 1H-indol-2-ylmethyl (2j), and 5-phenylthiophen-2-ylmethyl (2l) substituents
exhibited the highest potency against class I HDACs: HDAC1 IC50 39-61 nM, HDAC2 IC50 260-
690 nM, HDAC3 IC50 25-68 nM, and HDAC8 IC50 320-620 nM. The cytotoxicity of a
representative set of secondary and tertiary N-substituted 7-aminoheptanoic acid hydroxyamide-
based inhibitors against HT-29, SH-SY5Y, and MCF-7 cancer cells correlated with their inhibition
of HDAC1, 2, and 3 and was comparable to or better than that of SAHA (1). Compounds in this
series increased acetylation of histones H3 and H4 in a time-dependent manner. DMPK studies
indicated that secondary amine 2j is metabolically stable and has plasma and brain concentrations
>23- and >1.6-fold higher than the IC50 for class I HDACs, respectively. Overall, the secondary
and tertiary N-substituted 7-aminoheptanoic acid hydroxyamide-based inhibitors exhibit excellent
leadlike/druglike properties and therapeutic capacity for cancer applications.
KEYWORDS: Antitumor agents, Histone deacetylase, Inhibitors, Amines, Epigenetics.
2
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Introduction:
Histone deacetylases (HDACs) are key epigenetic regulators.[1] The zinc-dependent HDACs are
divided into three classes based on structure, sequence homology, and domain organization.[1]
Class I consists of HDACs 1, 2, 3, and 8, class II – HDACs 4, 5, 6, 7, 9, and 10, and class IV –
HDAC 11. Deacetylation of histone substrates results in an overall change in the post-translational
state of histones, known as the “histone code”.[2] The list of cellular events controlled by HDACs
has grown beyond DNA replication, DNA repair, chromatin remodeling, and gene transcription to
non-histone targets and noncoding mRNA, and it continues to expand.[3] Normal regulation of
these processes is compromised in a variety of diseases and conditions, and altered HDAC
expression/function has been shown to be a hallmark of many cancers and neurodegenerative and
inflammatory diseases.[3b, 4] Because of the roles HDACs play in these diseases, they have emerged
as potential therapeutic targets. The FDA has approved pan-HDAC inhibitors Zolinza (SAHA),
Beleodaq (belinostat/PXD101), Farydak (panobinostat), and class I selective HDAC inhibitor
Istodax (romidepsin) to treat peripheral or cutaneous T-cell lymphoma and multiple myeloma.[5]
While HDAC inhibition is a promising therapeutic strategy, HDACs play essential roles in normal
cellular function.[6] It has been hypothesized that HDAC isoform selective inhibitors would have
improved efficacy and minimal adverse effects; thus, isoform selective compounds have been
developed (for review see ref. [7]). In general, inhibition of class I HDAC isoforms and in some
cases HDAC6, a class II isoform, is associated with anti-cancer activity.[8] For instance, it has been
shown that overexpression of class I HDACs is correlated with a decrease in overall survival in
prostate,[9] colon,[10] breast,[11] lung,[12] liver,[13] gastric,[14] and neuronal[15] cancers and that class I
HDAC isoforms play predominant roles in epigenetic repression of key tumor suppressor genes
and genes involved in DNA damage repair in several tumor types.[8a, 16] Despite significant
3
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
progress, many aspects of HDAC biology are not well characterized or understood. Among them
are the role of individual HDAC isoforms in disease, the actual engagement of HDAC isoforms
with inhibitors in vivo, or the compensatory action of one isoform for another when one isoform
or a set of isoforms is inhibited. Likely for these reasons, the therapeutic application of HDAC
inhibitors remains somewhat limited. Therefore, discovery of novel potent class I HDAC
inhibitors, especially those with superior medicinal chemistry and anticancer properties, remains
an important task for development of epigenetics-based therapeutics.
One of the key features of the binding site in the class I and II HDAC isoforms is an aspartic amino
acid Asp104 (in HDAC2, different number in other HDACs) located at the gorge region of the
binding site. The acidic side chain of this well conserved residue may be considered a “hot-spot”
in the HDAC binding site. Being relatively solvent exposed, Asp104 is expected to be
deprotonated, yet its protonation state and its precise role in binding inhibitors and histone
substrates remains a matter of debate.[17] To probe the interaction between the Asp104 “hot spot”
and ligands, we designed and synthesized a series of compounds with an aliphatic amino group as
a part of their surface binding group (SBG, Figure 1). Although several HDAC inhibitor scaffolds
containing a basic nitrogen have been explored,[18] there is no systematic investigation of the effect
secondary and tertiary amines in the SBG may have on potency and HDAC isoform selectivity.
To minimize the effect of the remaining portion of the ligands on the structure activity relationship
(SAR), we focused our studies on compounds with the same linker and zinc binding group (ZBG,
Figure 1). In our recent publication, we have already determined that the linker consisting of six
methylene groups results in inhibitors more potent than those with a shorter linker.[19] The
hydroxamic acid and ortho-aminoanilide moieties are the two ZBGs (Figure 1) most commonly
used for HDAC inhibitor design. Hydroxamic acid is also present in the FDA approved HDAC
4
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
inhibitors. The advantages and disadvantages of both ZBGs remain an area of active
investigation.[20] Unlike the hydroxamic acid ZBG, however, the ortho-aminoanilide ZBG is
known to skew inhibition toward HDAC1-3 isoforms.[21] To minimize potential bias of ortho-
aminoanilide ZBG on SAR, we centered our efforts on hydroxamic acid-based compounds. In this
paper, we report the design, synthesis, docking, inhibition of recombinant class I and cellular
HDAC isoforms, biological evaluation in cellular models of cancer, and preliminary drug
metabolism and pharmacokinetic studies of a novel series of HDAC inhibitors containing either
secondary or tertiary aliphatic amino group as their SBG.
Figure 1. The FDA-approved inhibitor SAHA (1) and general structure of amines 2 and 3.
5
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Results and Discussion
The synthesis of all the compounds is shown in Scheme 1 and the structures of the substituents R1
and R2 are shown in Table 1. The synthesis of the secondary and tertiary amine-based HDAC
inhibitors is based on a reductive amination procedure, which has been described by us and
others.[22] A small library of commercially available aromatic aldehydes 1a-m were reacted with
methyl 7-aminoheptanoate. The resulting secondary and tertiary amines 6a-m and 7a-m,
respectively, were isolated and purified. The subsequent treatment of 6a-m and 7a-m with NH2OH
in MeOH gave the target hydroxamic acids 2a-m and 3a-m (Scheme 1).
Reagents and conditions: (a) Et3N (2.5 eq), STAB (2.5 eq), CH2Cl2, rt, 5 h, (yields: 6a-m
50-60%, 7a-m 15-30%); (b) NH2OH·HCl (200 eq), KOH (205 eq), MeOH, 0 °C−rt, 3 h
(yields: 40-80%).
Scheme 1. General synthetic scheme for secondary and tertiary amine-based HDAC inhibitors.
Considering the importance of inhibition of class I isoforms in cancer, we mainly focused on
testing the inhibitory activity against HDAC isoforms 1, 2, 3, and 8. A representative set of amine-
based inhibitors was also tested in cells for inhibition of acetylation of α-tubulin, a well validated
6
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
cellular target of HDAC6. The IC50 values of amines 2a-m and 3a-m for deacetylase activity of
class I HDACs are shown in Table 1. They were determined using a competitive fluorescence-
based assay similar to that previously reported by us.[23] Briefly, the inhibition of HDAC1, 2, and
3 was measured using the fluorescent HDAC substrate Boc-L-Lys(Ac)-AMC and commercially
available recombinant human HDAC1, 2, and 3 expressed in baculovirus expression system,
whereas the inhibition of HDAC8 was measured using the commercially available HDAC
substrate and purified recombinant human HDAC8 from Escherichia coli.[24]
7
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Table 1. Inhibition profile of secondary and tertiary amine-based HDAC inhibitors.
IC50 (nM)[a]
Compd
R1
R2
HDAC1
HDAC2
HDAC3
HDAC8
1
-
22.0 ± 5.0
200 ± 14
21 ± 8.5
210 ± 15
2a
H
H
H
340 ± 3.5
3100 ± 790
430 ± 380
1800 ± 130
3a
2b
3b
1100 ± 59
220 ± 25
2000 ± 92
2700 ± 670
2900 ± 60
4000 ± 210
1100 ± 120
1000 ± 200
1400 ± 61
4000 ± 370
2100 ± 50
4900 ± 550
2c
3c
2d
220 ± 29
1500 ± 110
240 ± 10.0
3600 ± 37
190 ± 34
1700 ± 120
8300 ± 1100
2800 ± 100
3900 ± 100 10000 ± 320
H
H
140 ± 12
790 ± 38
3d
2e
3400 ± 97 15000 ±1800 4700 ± 84
9800 ± 310
1800 ± 76
210 ± 18
270 ± 11
1800 ± 98
180 ± 5.0
290 ± 26
3e
1100 ± 560
2100 ± 83
2f
3f
H
H
430 ± 22
480 ± 65
39 ± 3.00
3200 ± 130
4000 ± 370
320 ± 13
310 ± 13
480 ± 29
68 ± 2.0
1600 ± 380
1400 ± 160
320 ± 26
2g
8
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
3g
2h
1100 ± 47
130 ± 10
4200 ± 250
840 ± 94
950 ± 49
170 ± 9.0
5400 ± 470
270 ± 44
H
H
3h
2i
500 ± 45
250 ± 19
680 ± 66
4200 ± 52
1800 ± 140
3700 ± 160
760 ± 26
120 ± 7.0
4200 ± 580
720 ± 20
3i
2900 ± 240
25 ± 2.0
2100 ± 80
620 ± 42
2j[b]
3j
H
61 ± 5.9
340 ± 20
260 ± 15
3700 ± 98
1300 ± 160
4600 ± 100
2k
3k
2l
H
H
130 ± 20
1300 ± 160
48 ± 5.6
620 ± 28
3700 ± 98
690 ± 31
160 ± 12
440 ± 11
38 ± 2.0
800 ± 30
6400 ± 70
550 ± 94
3l
840 ± 35
3000 ± 75
1100 ± 123
1500 ± 91
4100 ± 30
2m
H
1500 ± 110
3000 ± 140
2500 ± 350
3m
1700 ± 70
6700 ± 1100
1600 ± 92
25000 ± 3700
[a] IC50 values are expressed as mean ± standard deviation of at least two independent experiments.
The numbers are rounded to two significant figures. [b] HDAC6 IC50 = 67 ± 0.23 nM
9
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
We found that replacement of the amide moiety in SAHA (1) to an amino group in amine 2a led
to a 8- to 40-fold decrease in potency against all class 1 HDACs to 340, 3100, 430, and 1800 nM,
respectively (Table 1). To investigate how substitution of the aromatic ring in 2a with electron
withdrawing groups (EWGs) and/or electron donating groups (EDGs) would affect potency, we
synthesized and screened secondary amines 2b-e.
Overall, the potency and the pattern of HDAC isoform inhibition were only moderately affected
by addition of EWGs and/or EDGs compared to those of 2a. For amines 2b-e, the potency for
HDAC1 ranged from 140-220 nM, HDAC2 – 790-2900 nM, HDAC3 – 180-1000 nM, and HDAC8
– 1700-2800 nM. No particular EWG- or EDG-dependent trend was observed. Introduction of
these substituents in amines 2b-e generally resulted in a moderate improvement in potency against
HDAC1, 2, and 3 and less than 30% change in potency against HDAC8. Specifically, introduction
of p-NO2 in 2b, p-fluoro in 2c, p-fluoro and m-CH3 in 2d, and m,p-OCH3 in 2e improved HDAC1
potency of these compounds from 340 nM for amine 2a to 220, 220, 140, and 210 nM, respectively.
HDAC2 potency also showed only a moderate improvement from 3100 nM for 2a to 2900, 1500,
790, and 1800 nM for 2b-e, respectively. For compound 2b, potency against HDAC3 showed a
2.3-fold decrease, whereas for compounds 2c-e potency improved to 240, 190, and 180 nM,
respectively. Potency against HDAC8 remained in the single digit micromolar range – 2100, 1700,
2800, and 1800 nM for 2b-e, respectively.
The introduction of EWG and EDG substituents in amines 2b-e had either slightly improved class
I HDAC isoform selectivity or had no effect. The presence of the p-nitro group, a strong EWG,
moderately improved the selectivity of 2b towards HDAC1, whereas a combination of p-fluoro
and m-methyl substituents, a weak EDG and an EWG, respectively, made compound 2d more
10
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
selective for HDAC1, 2, and 3 over HDAC8. The selectivity of 2c and 2e remained relatively
comparable with 2a; that is, selective for HDAC1 and 3 over HDAC2 and 8.
Next, we explored how changes in the size, lipophilicity, and polarity of the SBG substituents
would affect activity of the ligands (Table 1). The pyridine ring in amine 2f had little effect on
HDAC inhibition compared to that of amine 2a, 430, 3200, 310, and 1600 nM for HDAC1, 2, 3,
and 8, respectively. Replacement of the phenyl ring in 2a with fused bicyclic moieties naphthyl,
methylenedioxophenyl, N-substituted indole, and indole in 2g-j, respectively, had by far the most
robust effect on improvement of the potency for all class I HDAC isoforms. Potency of compounds
2g-j for HDAC1 was 39, 130, 250, and 61 nM, HDAC2 – 320, 840, 1800, and 260 nM, HDAC3 –
68, 170, 120, and 25 nM, and HDAC8 – 320, 270, 720, and 620 nM, respectively. The selectivity
profile for 2g-j was similar to amine 2a (Table 2). Replacement of the phenyl ring in 2a with bi-
aryl substituents in amines 2k and 2l had an effect similar to that found in the amines with bicyclic
substituents. These compounds were superior to amine 2a and maintained the overall HDAC
isoform selectivity profile. Specifically, placement of a pyridine ring in the para position of the
phenyl group in amine 2k resulted in an improvement in IC50 for all class I HDACs in comparison
to compound 2a (Tables 1 and 2), 130, 620, 160, and 800 nM for HDAC1, 2, 3, and 8, respectively.
Introduction of a bicyclic ring system of thiophene had resulted in potent and selective HDAC1
and 3 inhibitor 2l (Table 1 and 2). The values of IC50 against HDAC1 and 3 for 2l were 48 and 38
nM, respectively. The selectivity profile of 2l had shown 11 to 14-fold difference in inhibition
between HDAC1 and 3 versus HDAC2 and 8 (Table 2). Introduction of a longer SBG in amine
2m resulted in a substantial loss of activity - 1500, 3000, 1500, and 2500 nM, for HDAC1, 2, 3,
and 8, respectively.
11
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Table 2. Selectivity profile of secondary and tertiary amine-based HDAC inhibitors.
Selectivity[a]
Compd
HDAC2/HDAC1
HDAC3/HDAC1
HDAC8/HDAC1
2a
3a
2b
3b
2c
3c
2d
3d
2e
3e
2f
9.1
2.5
13
1.3
1
5.3
3.6
9.5
2.5
7.7
2.1
20
4.5
0.7
1.1
0.92
1.4
1.4
0.86
1.1
0.72
1
2
6.8
2.6
5.6
4.4
8.6
4.1
7.4
8.3
8.2
3.8
6.5
8.4
7.2
5.4
4.2
11
2.9
8.6
7.8
3.7
2.9
8.2
4.9
2.1
8.4
2.9
3.1
10
3f
2g
3g
2h
3h
2i
1.7
0.86
1.3
1.5
0.48
4.3
0.41
3.8
1.2
0.34
0.8
1.3
1
3i
2j
3j
14
2k
3k
2l
4.8
2.8
14
6.2
4.9
11
3l
3.6
2
4.9
1.7
15
2m
3m
3.9
0.94
[a] Selectivity ratios are calculated by dividing the HDAC1, 2, 3, and 8 IC50 of an amine-based
inhibitor by HDAC1 IC50 of the same inhibitor. The numbers are rounded to two significant
figures.
12
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
We also synthesized the corresponding tertiary amines and found that, with few exceptions, their
potency varied from single to double digit micromolar. An introduction of an additional tertiary
substituent in amines 3a-d and 3h-l had resulted in less potent inhibitors than the corresponding
secondary amines (Table 1). The effect of the size, lipophilicity, and polarity was rather
unpronounced compared to that of the secondary amines. In the case of 3,4-dimethoxy and 3-
pyridine substituents, however, additional SBGs in amines 3e and 3f did not cause any significant
changes in their potency and resulted in an inhibitory profile similar to that of the corresponding
secondary amines 2e and 2f (Table 1). In 3i and 3j, the additional methylenedioxophenyl, N-
substituted indole and indole moieties have resulted in compounds more potent against HDAC1 in
comparison with the corresponding secondary amines (Table 1 and Table 2). Potency of
compounds 3i and 3j for HDAC1 was 680 and 340 nM, HDAC2 – 3700 nM, HDAC3 – 2900 and
1300 nM, and HDAC8 – 2100 and 4600 nM, respectively. In 3m, the additional long linear
substituent resulted in almost no changes in potency against HDAC1 and 3 and a 2.2- and 10-fold
decrease in potency against HDAC2 and 8, respectively, compared to corresponding secondary
amine 2m. Among the tertiary amines, 3e was the most potent inhibitor with IC50 of 270, 1100,
290, and 2100 nM against HDAC1, 2, 3, and 8, respectively. In contrast to secondary amine 2l that
was more selective for HDAC1 and 3, tertiary amines 3i and 3j were more selective toward
HDAC1 and tertiary amine 3k was more selective toward HDAC3 (Table 2).
We have previously demonstrated that tertiary amine-based HDAC inhibitors can be converted to
photoreactive probes P1 and P2 (Figure 2A) by introducing a photoreactive 3-azido-5-
azidomethylene benzyl moiety as one of the substituents at the basic nitrogen atom and the other
substituents are either an indole group (P1) or a 5-(4-tert-butoxycarbonylaminophenyl) isoxazole
group (P2).[25] We used these probes as nanorulers to determine the distance between the catalytic
13
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
site of HDAC3 and its co-activator silencing mediator of retinoid and thyroid hormone receptors
(SMRT-DAD). Given the observed difference in potency of the SBGs in pairs 2j/3j and 2m/3m
(Table 1), we sought to determine the effect of placing two different groups on the inhibitory
profile and if this modification affects their binding and/or suitability for photolabeling
experiments against other class I HDACs. Probes P1 and P2 displayed moderate potency (Figure
2A) against class I HDAC isoform, which is in agreement with the potency for the other tertiary
amines reported here. Potency of P1 and P2 for HDAC1 was 2300 and 1200 nM, HDAC2 – 4700
and 5600 nM, HDAC3 – 890 and 590 nM, and HDAC8 – 6000 and 14000 nM, respectively (Figure
2A). The introduction of 3-azido-5-azidomethylene benzyl moiety in P1 and P2 has resulted in a
better inhibitory profile toward HDAC3 with at least 2-fold increase in potency between HDAC1
and 3 and 5-23 fold increase between HDAC3 and HDAC2 and 8. These results suggest that
introduction of any bulky tertiary substituent leads to an overall lower activity. The improved
potency of P1 and P2 suggests that further improvement of potency and selectivity of tertiary
amines 3 can be achieved upon additional SAR studies but is unlikely to be substantial. Next, we
performed the photolabeling experiments with probes P1 and P2. In this type of photoreactive
probes, the aromatic azide is used to generate a reactive nitrene upon UV irradiation thereby
forming covalent adducts with HDACs, whereas the aliphatic azide reacts with a reporter tag, e.g.
the biotin-containing tag shown in Figure 2A, via a “click-chemistry” reaction.[23, 25-26] At a fixed
concentration of 8.5 µM, both P1 and P2 can label recombinant HDAC1 and 8 (Figure 2B, 2C),
whereas the labeling of HDAC3 was demonstrated previously.[25] Only a marginal and likely non-
specific biotinylation of HDAC1 and 8 is observed in the experiments where the proteins were
preincubated with 42.5 µM of trichostatin A (TSA), a non-selective HDAC inhibitor. Overall, these
experiments demonstrate that P1 and P2 can be used as photolabeling probes against all class I
14
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
HDACs and, hence, represent additional tools for future target engagement and target
identification experiments in live cells for class I HDACs.[27]
15
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Figure 2. Photolabeling experiments against HDAC1 and HDAC8 using photolabeling probe
1 and 2. A) Chemical structures and activity profile of P1, and P2. B) HDAC1 (1.3 µ ) was
incubated with either photolabeling probes P1 (8.5 µ ) and P2 or no probes control C in
presence/absence of TSA (42.5 µ ) for 2-3 h in the dark, followed by UV irradiation for 3 min to
M
M
M
activate the aromatic azido group to form a covalent bond with nearby reactive amino acids side
chains, then the click chemistry reaction was initiated between benzyl azido group and biotin-
alkyne tag (50 µ
conjugated horse radish peroxidase (Strep-HRP). C) Similar to panel B using HDAC8 at a final
concentration of 1.7 µ . IC50 values are expressed as mean ± standard deviation of at least two
M). After 1 h, protein samples were analyzed via Western blots using streptavidin
M
independent experiments. The numbers are rounded to two significant figures. Equal loading of
protein samples was validated using anti-HDAC1 antibody or coomassie staining.
16
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
To gain additional structural insights into the SAR, amines 2 and 3 were docked to HDAC2 (PDB:
4LXZ[28]) using Molecular Operating Environment (MOE) software.[29] The top docking poses of
two representative amines 2j and 3e are shown in Figure 3A and 3B, respectively. A 2D map of
the interactions of these ligands with HDAC2 is shown in Figure 4. An analysis of the docking
poses shows that the SBG of the secondary amines occupies one of the hydrophobic grooves and
forms a salt bridge between the charged secondary amino group of the ligands and Asp104 of
HDAC2. Amines with basic nitrogen atoms in the SBG may form an additional polar interaction
with Glu103 similar to that shown for 2j in Figure 3A and 4A. The exact placement of the aryl
substituents in the binding site depends on their size, shape, and electronic properties. It tends to
gravitate to the poses with the largest area of contact with the hydrophobic portions of the binding
site and, whenever possible, a salt bridge between the charged amino group and the ionized side
chain of Asp104. Considering our previous studies and availability of multiple conformations with
similar scores for the docked amine-based inhibitors,[30] binding of these compounds as an
ensemble of poses rather than a single pose cannot be excluded. The former would also result in a
smaller loss in entropy and, hence, better binding. For systems similar to amines 2 and 3 bound to
HDACs, salt bridges were shown to contribute on average 12-21 kJ/mol to the protein stabilization
energy,[31] which is notably more than the typical 5.0 ± 2.5 kJ/mol contribution of a hydrogen bond
to the binding.[32] Despite the possible advantage of having a salt bridge between the protonated
amino group of the ligands and deprotonated side chain of Asp104 compared to a hydrogen bond
between these groups in a neutral form, it is unclear if this is the case. The fact that all the secondary
amines 2, even a nearly identical to 1 amine 2a, are less potent than 1 suggests that the charged
amino group does not gain free energy of binding comparable to that of 1 likely due to a higher
overall solvation-desolvation penalty. The docking of tertiary amines 3 shows that in all the poses
17
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
both the substituents share a rather narrow gorge of the binding site, which likely results in a large
entropic loss. Only one methylene spacer between the amino group and the aryl group offers a
very limited set of conformations, if any, in which both the substituents can form enthalpically
favorable interactions with the binding site. Moreover, tertiary amines 3 are limited in their choice
between a binding pose where there is a salt-bridge with Asp104 and marginal interaction between
the aromatic substituents and the lipophilic portion of the binding site as shown for 3e (Figures
3B and 4B) and a conformation where the aromatic substituents (or at least one of them) form
pronounced interaction with the hydrophobic area of the binding site whereas the distance between
the negative side chains of Asp104 and Glu103 and positively charged tertiary amine is extended
to at least 5-6 Å. Lacking additional bulky tertiary substituent, secondary amines 2 are much less
restricted in their poses and maintain both these interactions with the binding site simultaneously.
These observations suggest that both the higher entropic loss and the smaller enthalpic gain upon
binding of tertiary amines 3 are likely the reasons they are less potent than corresponding
secondary amines 2.
18
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Figure 3. (A) Docked pose of amine 2j in the binding site of HDAC2 (PDB: 4LXZ) and (B) same
for amine 3e. The binding site surface is shown as a surface colored with lipophilic potential,
green- lipophilic, purple – hydrophilic.
Figure 4. Protein-ligand interaction between HDAC2 and (A) amine 2j and (B) amine 2k in PDB:
4LXZ. The 2D depiction of protein-ligand interactions in panels A and B is described in ref.[33]
19
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
In silico druglike properties of amine-based HDAC inhibitors were calculated in MOE and
included water/octanol partition coefficient SlogP and water/octanol distribution coefficient at pH
7 logD as descriptors of lipophilicity, solubility logS, and topological surface area TPSA. The
lipophilic ligand efficiency LLE was calculated as reported by Ryckmans et al[34] in
QikProp/Schrödinger software.[35] An analysis of the calculated logP, logS, TPSA, MW, logD, and
LLE given in Table 3 indicates that the secondary and tertiary amines are generally
leadlike/druglike and are excellent starting point for further drug discovery efforts.[36] Low
molecular weight, TPSA below 90 Å, presence of a basic aliphatic nitrogen atom, logD in the
range of 0-3, and a number of nitrogen and oxygen atoms below 5 suggest that these compounds
have high probability to be brain-blood barrier (BBB) permeable.[37] The majority of potent
secondary amines 2 are characterized by LLE above 4 and calculated logP between 2 and 3,
indicating that these compounds are likely to have acceptable ADME properties.[38] Additionally,
in silico evaluation of secondary and tertiary amines activity against hERG potassium channel, a
predictor of QT prolongation and cardiac toxicity,[39] were performed using QikProp/Schrödinger
software (Table 3).[35] In all the cases, secondary amines 2 were found to be less potent against
hERG than the corresponding amines 3. With few exceptions, the secondary and tertiary amine-
based inhibitors displayed acceptable (greater than -5) predicted logIC50 for hERG activity.
20
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Table 3. Physicochemical properties of secondary and tertiary amine-based HDAC inhibitors.
Physicochemical Properties[a]
LLE
HD1
LLE
HD2
LLE
HD3
LLE
HD8
Compd SlogP
logS
TPSA
MW
logD
loghERG
1
2a
3a
2b
3b
2c
3c
2d
3d
2e
3e
2f
2.5
2.5
4.7
2.4
4.5
2.6
5.0
2.9
5.6
2.5
4.7
1.9
3.5
3.7
7.0
2.2
4.1
4.8
9.2
3.0
5.6
3.6
6.8
4.2
8.1
4.5
8.7
-2.9
-2.2
-4.1
-3.0
-5.7
-2.5
-4.7
-2.7
-5.0
-2.3
-4.3
-0.95
-1.6
-4.1
-7.9
-2.2
-4.0
-4.4
-8.4
-2.7
-5.0
-3.3
-6.4
-4.4
-8.5
-4.7
-9.0
78
61
53
110
140
61
53
61
53
80
89
74
78
61
53
80
89
76
81
77
84
74
78
61
53
130
180
260
250
340
300
430
270
380
280
410
310
460
250
340
300
440
290
430
410
660
290
420
330
490
330
500
430
700
1.9
-0.73
1.8
-4.3
-5.0
-5.2
-4.9
-5.7
-4.8
-5.5
-4.8
-5.4
-4.8
-5.6
-4.7
-5.5
-5.6
-7.0
-4.9
-5.8
-6.1
-7.2
-5.3
-6.6
-6.0
-7.9
-5.8
-7.7
-6.3
-7.9
5.2
4.0
4.2
3.0
5.2
3.9
4.2
3.2
1.3
0.89
3.1
1.3
0.72
3.3
-0.40
2.4
4.3
3.6
1.2
0.90
3.2
1.4
0.82
3.1
-0.51
2.2
4.0
4.0
0.45
3.9
0.044
3.2
0.49
3.8
0.13
2.6
-0.30
2.6
-0.10
4.2
-0.75
3.2
-0.25
4.2
-0.56
3.2
-0.82
1.5
1.9
1.2
1.8
0.97
3.9
-0.73
1.2
4.5
3.6
4.6
3f
2.9
1.9
2.9
2.4
2g
3g
2h
3h
2i
0.42
4.0
3.8
2.8
3.5
2.8
-1.0
4.7
-1.6
3.8
-0.96
4.5
-1.7
4.3
-1.3
0.64
2.2
2.2
1.2
2.0
1.2
1.8
0.97
-3.8
3.6
2.1
1.4
3i
7.4
-3.1
4.2
-3.7
4.6
-3.7
3.2
2j
0.087
3.20
0.31
3.8
3j
0.83
3.3
-0.21
2.6
0.24
3.2
-0.30
2.5
2k
3k
2l
-0.92
3.1
-1.4
1.9
-0.45
3.2
-1.6
2.0
1.7
3l
6.4
-2.1
1.3
1.9
3.2
2.0
2m
3m
1.1
1.0
1.3
1.1
5.0
-2.9
-3.5
-2.9
-4.1
[a] Physicochemical properties were calculated with MOE and QikProp/Schrödinger software. The
numbers are rounded to two significant figures. Detailed description of these parameters can be
found in the method section in supplementary information. LLE HD1, 2, 3, and 8 stands for
lipophilic ligand efficiency calculated for HDAC1, HDAC2, HDAC3, and HDAC8, respectively.
21
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Next, a representative set of seven amine-based inhibitors 2g, 2h, 2j-l, 3e, and 3f and the parent
compound 1 were tested for antiproliferative activities against three cancer cell lines of human
origin: colorectal adenocarcinoma HT-29, neuroblastoma SH-SY5Y, and breast adenocarcinoma
MFC-7, using an alamarBlue assay.[40] The EC50 against HT-29 and SH-SY5Y cells and percent
growth inhibition at 10 µM against MCF-7 cells are shown in Table 4. The EC50 were measured
at 24 and 48 hours. In case of HT-29 cells, the EC50 at 24 hours for 1 and all the amines tested
were above 50 µM, except for compounds 2k and 2l exhibiting EC50 of 35 µM and 24 µM,
respectively. At 48 h, EC50 against HT-29 cells for 1 and amines 3e, 2g, 2h, and 2j-l were in the
range between 1.1 µM and 4.2 µM with compound 2g being the most potent with an EC50 of 1.1
µM. Amine 3f was ineffective against HT-29 cells even at 48 h time point. In case of SH-SY5Y
cells, the EC50 at 24 h for 1 and amines 2g, 2h, 2j, 3e, and 3f were above 50 µM. Amines 2k and
2l displayed EC50 of 13 and 15 µM, respectively. At 48 h, EC50 for 1 and all the amines tested
ranged between 1.2 to 23 µM. Potency of amines 2g and 2j was superior to that of 1, 1.2 and 1.3
µM, respectively. Except for amine 3f that exhibited EC50 of only 23 µM, potency of other amines
was either comparable or somewhat lower than that of 1. In case of MCF-7, the calculated
percentage of inhibition at 24 h for 10 µM of 1 and amines 2j-l and 3e was 48%, 61%, 66%, 63%,
and 55%, respectively, whereas amines 2g, 2h, and 3f displayed less than 25% of inhibition. At 48
h, the percent of inhibition by amines 2g, 2h, 2j-l, and 3e was above 46%. Amines 2j and 2l, both
with 71% of inhibition, were found to be slightly more potent than 1, which exhibited 69% of
inhibition. Amine 3f, on the other hand, was almost inactive and displayed only an 8.7% inhibition
of MCF-7 cells growth. These results show that secondary amine-based HDAC inhibitors have
comparable or in some cases better cytotoxicity profile than that of 1.
22
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Table 4. Cytotoxicity of 1 and amine-based HDAC inhibitors 2g, 2h, 2j-l, 3e, 3f against HT-29,
SH-SY5Y, and MCF-7 cell lines.
HT-29[a]
48 h
SH-SY5Y[a]
24 h
MCF-7[b]
48 h
Compd
24 h
48 h
24 h
1
> 50
> 50
> 50
1.0 ± 0.11
1.1 ± 0.11
2.2 ± 0.68
2.1 ± 0.17
2.3 ± 0.21
2.5 ± 0.65
4.2 ± 0.37
48 ± 1.4
> 50
> 50
> 50
1.5 -± 0.1
1.2 ± 0.22
2.4 ± 0.32
1.3 ± 0.39
1.5 ± 0.12
6.1 ± 0.78
3.5 ± 0.16
23 ± 1.8
48%
19%
24%
61%
66%
63%
55%
NA[c]
68%
46%
56%
71%
63%
71%
60%
8.7%
2g
2h
2j
2k
2l
> 50
> 50
35 ± 4.0
24 ± 1.4
> 50
13 ± 1.6
15 ± 1.1
> 50
3e
3f
> 50
> 50
[a] EC50 values (µM) are expressed as mean ± standard deviation of four independent experiments
(n=4). [b] Growth inhibition percentage at 10 µM. [c] NA = no inhibition at 10 µM.
23
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
To enable analysis of the correlation between the activity of these compounds against recombinant
enzymes and the cytotoxicity data, we calculated the correlation coefficients (R) between all the
IC50 and EC50 at 48 h for the compounds in Table 4. The complete data are given in
Supplementary Figure 1. The IC50 values for HDAC1, 2, and 3 are highly correlative, with
correlation coefficient R ranging between 0.91 and 0.99. Considering very high homology between
the sequences of HDAC1, 2, and 3, such high correlation observed between the potencies against
these enzymes appears to be reasonable. Correlation of IC50 for HDAC1, 2, and 3 with those of
HDAC8, a less homologous isoform, is lower, with R of 0.75, 0.53, and 0.73, respectively. A
similar correlation analysis of IC50 values for each individual isoform and EC50 against HT-29 and
SH-SY5Y cells shows a strong correlation between inhibition of HDAC1, 2, and 3 isoforms and
cytotoxicity, with R ranging between 0.81 and 0.98. The R for the correlation between EC50 against
HT-29 and SH-SY5Y cells and IC50 for HDAC8 is 0.43 for both cell lines. These data suggest that
the cytotoxicity stems largely from inhibition of either individual HDAC isoform 1, 2, and 3 or
their combinations. The correlation coefficient between cytotoxicity for both cell lines and activity
against HDAC2 was found to be somewhat higher, 0.98, compared to that for the other
combinations of the isoforms and the cell lines. Strong intercorrelation between IC50 values for
HDAC1, 2, and 3 does not allow to identify a particular isoform(s) primarily responsible for
cytotoxicity. Removal of amines 2l and 3f, two compounds with very poor EC50 values that can
artificially improve correlation, has resulted in generally similar correlation. Interestingly, it also
resulted in substantial improvement in correlation with IC50 for HDAC8, 0.93 and 0.78 for HT-29
and SH-SY5Y, respectively. The presence of thiophene ring, which is a known metabolic liability,
in amine 2l and differences in bioenergetics between SH-SY5Y and HT-29 cells may account for
lower than expected (based on its HDAC inhibitory profile) cytotoxicity of amine 2l in SH-SY5Y
24
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
cells. Alternatively, poor cell permeability or precipitation of 2l and 3f, although the latter was not
observed upon visual inspection, may also affect their potency in cell-based assays. To determine
whether this may be the case, we compared logP, logS, and TPSA parameters for all compounds
in this series (Table 3). We found that for both 2l and 3f these parameters were similar to those of
the other compounds tested for cell-based cytotoxicity, with compound 2l having the lowest
calculated solubility logS of -4.41. It suggests that at least in case of 2l solubility may potentially
affect its activity in cell-based assays.
Next, we sought to validate acetylation of histones 3 (H3) and 4 (H4) as the target for amines 2e,
2g, 2h, 2j, 2k, 3f, and 3k in HT-29 and SH-SY5Y cells by Western blot (Figure 5 and 6).
Compound 1 was used as a positive control. In HT-29 cells, the acetylation was measured at
preincubation times of 6 and 24 hours and at a concentration of 5 µM for all the compounds. These
preincubation times and the subtoxic dose for the inhibitors were selected to ensure that the effect
of inhibition is mediated by engaging the target while most cells are still alive. At 6 h, compound
1 and amines 2e, 2g, 2h, 2j, 2k, 3f, and 3k increased the acetylation level of H3 and only compound
2g was able to significantly increase acetylation of H4 (Figure 5A). At 24 h, a time-dependent
increase in acetylation of H3 and H4 was observed for amines 2e, 2g, 2h, and 2j, and in H4 only
for 2k (Figure 5B). Compounds 1, 3f, and 3k were unable to cause any significant time-dependent
increase in acetyl H3 and acetyl H4. The inability of 1 to further increase acetylation of H3 and
H4 at 24 h in HT-29 cells prompted us to investigate the acetylation patterns in SH-SY5Y cells
under same conditions (Figure 6). At 6 h, the acetylation of both H3 and H4 was increased by
compound 1 and amines 2e, 2g, 2h, 2j, 2k, 3f, and 3k (Figure 6A). At 24 h, a time-dependent
increase in acetylation of H3 and H4 was observed for 2e, 2g, 2h, 2j, 2k, 3k, and 1, except for 3f
(Figure 6B). Overall, the ability of compounds to increase acetylation of H3 and H4 supports the
25
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
correlation between the IC50 values against HDAC1, 2, and 3 and the cytotoxic effects in HT-29
and SH-SY5Y cells (Supplementary Figure 1B), which is consistent with our previous
19, 41]
observations and those from other laboratories.[4d,
The differences in the global
hyperacetylation state in H3 and H4 at 24 h time point in HT-29 and SH-SY5Y cells in response
to the treatment with 1 indicate that its cytotoxic effect may be mediated via cell type-dependent
mechanisms that may involve non-histone targets as well. In fact, multiple modes of action of 1 in
HT-29 and other colorectal cancer cell lines were observed by other groups.[42] This finding
warrants further investigation into the mode of action of 1 and other HDAC inhibitors in different
cell lines for additional target identification.
26
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Figure 5. Analysis of histone H3 and H4 total acetylation status in HT-29 cells by Western
blotting. HT-29 cells were treated with either DMSO (V), 5 µ of 1, or 5 µ of 2e, 2g, 2h, 2j, 2k,
M
M
3f, or 3k at A) 6 h and B) 24 h. One-way ANOVA revealed significant increase in acetylation of
H3 and H4. The data is plotted as the average of at least 2 independent experiments +/- SD. (***,
p < 0.001; **, p < 0.01; *, p < 0.05; ns, statistically nonsignificant).
27
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
Figure 6. Analysis of histone H3 and H4 total acetylation status in SH-SH5Y by Western
blotting. SH-SH5Y cells were treated with either DMSO (V), 5 µ of 1, or 5 µ of 2e, 2g, 2h, 2j,
M
M
2k, 3f, or 3k at A) 6 h and B) 24 h. One-way ANOVA revealed significant increase in acetylation
of H3 and H4. The data is plotted as the average of at least 2 independent experiments +/- SD.
(***, p < 0.001; **, p < 0.01; *, p < 0.05; ns, statistically nonsignificant).
28
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700449
ChemMedChem
To determine if inhibition of class II HDACs, specifically HDAC6, by amines 2 and 3 can also
contribute to cytotoxicity, we determined their effect on acetylation of α-tubulin, a known
cytosolic substrate of HDAC6. At 6 h, compound 1 and amines 2e, 2g, 2h, 2j, 2k, and 3f increased
the acetylation level of α-tubulin in HT-29 cell lines, whereas amine 3k showed only a small and
not statistically significant increase (Figure 7A). At the same time point, amines 2g, 2h, 2j, 2k,
3f, and 3k and compound 1 increased the acetylation of α-tubulin in SH-SY5Y cells. Amine 2e
showed moderate but not a statistically significant increase. At 24 h, a time-dependent increase in
acetylation of α-tubulin in HT-29 cells was observed only for amine 2j and 1 (Figure 7B). Unlike
1, none of the amines tested induced statistically significant acetylation of α-tubulin in SH-SY5Y
cells at 24 h. To investigate this further, we determined the HDAC6 inhibitory activity of a
representative secondary amine 2j and found it to be a potent inhibitor of HDAC6 with an IC50 of
67 nM (Table 1, Supplementary Figure 2). Similar tertiary amine-based HDAC inhibitors have
been reported to be potent HDAC6 inhibitors as well.[19] These data suggest that the amine-based
HDAC inhibitors may inhibit HDAC6 transiently in cells, displaying a different time-dependent
inhibitory profile compared to 1. Considering similar structure and HDAC inhibitory profiles of
compound 1 and amines 2 and 3, the apparent time-dependent effect is likely due to the presence
of a basic aliphatic amino group in amines 2 and 3. The continuous acetylation of H3 and H4,
nuclear targets for class I HDACs, and the temporary hyperacetylation of α-tubulin, a cytosolic
target for HDAC6, suggest that the amine-based inhibitors 2 and 3 accumulate in a time-dependent
manner in the nucleus and possibly other organelles leading to a decrease in concentration in the
cytosol. Although further studies are needed to determine the origin of these observations, one of
the plausible explanations is a pKa/pH-dependent sequestration of amines into cellular
compartments/organelles that was previously reported for unrelated small molecules.[43]
29
This article is protected by copyright. All rights reserved.