Identification and Structure-Activity Relationship of HDAC6 Zinc-Finger Ubiquitin Binding Domain Inhibitors

Article abstract of DOI:10.1021/acs.jmedchem.8b00258

HDAC6 plays a central role in the recruitment of protein aggregates for lysosomal degradation and is a promising target for combination therapy with proteasome inhibitors in multiple myeloma. Pharmacologically displacing ubiquitin from the zinc-finger ubiquitin-binding domain (ZnF-UBD) of HDAC6 is an underexplored alternative to catalytic inhibition. Here, we present the discovery of an HDAC6 ZnF-UBD-focused chemical series and its progression from virtual screening hits to low micromolar inhibitors. A carboxylate mimicking the C-terminal extremity of ubiquitin, and an extended aromatic system stacking with W1182 and R1155, are necessary for activity. One of the compounds induced a conformational remodeling of the binding site where the primary binding pocket opens up onto a ligand-able secondary pocket that may be exploited to increase potency. The preliminary structure-activity relationship accompanied by nine crystal structures should enable further optimization into a chemical probe to investigate the merit of targeting the ZnF-UBD of HDAC6 in multiple myeloma and other diseases.

Full text of DOI:10.1021/acs.jmedchem.8b00258

Article
pubs.acs.org/jmc
Cite This: J. Med. Chem. XXXX, XXX, XXX−XXX
Identification and Structure−Activity Relationship of HDAC6 Zinc-
Finger Ubiquitin Binding Domain Inhibitors
Renato Ferreira de Freitas,†,# Rachel J. Harding,†,# Ivan Franzoni,‡ Mani Ravichandran,†
Mandeep K. Mann,† Hui Ouyang,† Mark Lautens,‡ Vijayaratnam Santhakumar,†
Cheryl H. Arrowsmith,†,§ and Matthieu Schapira*,†,∥
†Structural Genomics Consortium, University of Toronto, MaRS South Tower, Suite 700, 101 College Street, Toronto, Ontario M5G
1L7, Canada
‡Department of Chemistry, Davenport Chemical Laboratories, University of Toronto, 80 St. George Street, Toronto, Ontario M5S
3H6 Canada
§Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
∥Department of Pharmacology and Toxicology, University of Toronto, Toronto, Ontario M5S 1A8, Canada
S
* Supporting Information
ABSTRACT: HDAC6 plays a central role in the recruitment of
protein aggregates for lysosomal degradation and is a promising
target for combination therapy with proteasome inhibitors in
multiple myeloma. Pharmacologically displacing ubiquitin from the
zinc-finger ubiquitin-binding domain (ZnF-UBD) of HDAC6 is an
underexplored alternative to catalytic inhibition. Here, we present
the discovery of an HDAC6 ZnF-UBD-focused chemical series and
its progression from virtual screening hits to low micromolar
inhibitors. A carboxylate mimicking the C-terminal extremity of
ubiquitin, and an extended aromatic system stacking with W1182
and R1155, are necessary for activity. One of the compounds
induced a conformational remodeling of the binding site where the
primary binding pocket opens up onto a ligand-able secondary
pocket that may be exploited to increase potency. The preliminary structure−activity relationship accompanied by nine crystal
structures should enable further optimization into a chemical probe to investigate the merit of targeting the ZnF-UBD of
HDAC6 in multiple myeloma and other diseases.
ZnF-UBD domain of HDAC6, which would antagonize
HDAC6-mediated recruitment of ubiquitinated protein aggre-
gates to microtubules. We have recently developed a collection of
assays to enable the development of HDAC6 ZnF-UBD
chemical probes and reported fragment hits occupying its
binding pocket.17 Here, we describe the design and discovery of a
low micromolar HDAC6 ZnF-UBD inhibitor guided by virtual
screening and X-ray crystallography. The emerging structure−
activity relationship (SAR) defines several scaffolds as attractive
starting points for further elaboration into high affinity inhibitors.
INTRODUCTION
■
HDAC6 is the main cytoplasmic deacetylase in mammalian cells
and the only deacetylase containing a zinc finger ubiquitin-
binding domain (ZnF-UBD).1−3 Proteins deacetylated by
HDAC6 include α-tubulin,4 HSP90,5 and cortactin6 among
others.7,8 HDAC6 is involved in a diverse array of biological
processes, including cell motility,4 immune synapse formation,9
viral infection,10 and the degradation of misfolded proteins,11 and
has been associated with diseases.12 In particular, HDAC6
recruits polyubiquitinated protein aggregates via its ZnF-UBD
and loads misfolded proteins onto dynein for microtubule-
guided transport to the lysosome.11 This lysosomal degradation
pathway is an alternative mechanism to proteasomal degradation
for the clearance of misfolded proteins,11 and chemical inhibition
of HDAC6 represents a promising strategy to overcome
resistance to proteasome inhibitors in multiple myeloma and
lymphoma.13−15
RESULTS AND DISCUSSION
■
The crystal structure of the HDAC6 ZnF-UBD was previously
solved in the apo form (PDB: 3C5K) and in complex with an
ubiquitin C-terminal pentapeptide, RLRGG, (PDB: 3GV4) as
well as in complex with full-length ubiquitin (PDB: 3PHD).18
The C-terminal carboxylate of the ubiquitin peptide binds in a
deep cavity of HDAC6. Binding is stabilized by an extensive
Current HDAC6 inhibitors target the catalytic domain of the
enzyme, thereby inhibiting the deacetylation of microtubules,
and antagonizing the transport of protein aggregates to the
aggresome.16 An alternative mechanism of action is to target the
Received: February 15, 2018
© XXXX American Chemical Society
A
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 1. (a) Crystal structure (PDB: 3GV4) of the HDAC6 ZnF-UBD domain (gray ball and sticks) in complex with a ubiquitin C-terminal
pentapeptide, RLRGG (cyan ball and sticks); (b) Comparison of the ZnF-UBD of HDAC6 in the apo (PDB: 3C5K, magenta) and in the peptide bound
(orange) forms; (c) open (apo) conformation of HDAC6 ZnF-UBD; (d) closed (bound) conformation of HDAC6 ZnF-UBD. R1155 and Y1156 are
colored in orange.
network of hydrogen bonds with R1155, Y1156, Y1184, and
Y1189 as well as with conserved water molecules (Figure 1a).
Less conventional interactions such as an amide-π stacking19
between the C-terminal glycine and W1182, and two amide NH-
π bonds with W1143 and W1182, also contribute to binding.
Comparison of the apo and bound states reveals a significant
degree of plasticity at the binding pocket. Although the
conformation of most residues is conserved, important
conformational changes upon peptide binding are observed at
R1155 and Y1156 (Figure 1b). Both side chains relocate to form
hydrogen bonds with the peptide and bury the C-terminal
diglycine motif. Indeed, it was proposed that R1155 and Y1156
act as gatekeeper residues that move from an “open” (Figure 1c)
to a “closed” (Figure 1d) conformation upon ubiquitin binding.18
parallel, with and without a water molecule at the bottom of the
pocket (see Experimental Section for details), and 33
compounds (27 of which contained a carboxylate moiety)
were ordered and evaluated using a fluorescence polarization
(FP) peptide displacement assay. Six compounds displaced the
ubiquitin peptide at micromolar concentration (2, 13, 17, 18, 19,
20). We also ordered and tested 52 readily available chemical
analogues of the first hits, of which 6 were active (1, 8, 12, 14−
16). Overall, the hits could be grouped into two clusters, based
on the nature of the ring featuring the carboxylic chain: five-
membered ring (cluster 1, compounds 1−11) and six-membered
ring (cluster 2, compounds 12−20) (Table 1).
A preliminary SAR emerged from this ensemble of molecules.
Compound 1 was the most potent ZnF-UBD inhibitor, with an
IC50 of 5.1 μM. Common to compounds 1−6 was the presence
of a 3-(1,3-benzothiazol-2-yl) propanoic acid. Compound 2,
which is a substructure of compound 1, was weaker (IC50 = 180
μM) but had a better ligand efficiency (LE = 0.37). Analogues of
compound 2 featuring substituents on the ethylene linker (3−6)
were inactive. Replacing the propanoic acid in 2 with a butanoic
STRUCTURE−ACTIVITY RELATIONSHIP
To probe the chemical tractability of the ZnF-UBD, a virtual
library of ∼1.3 million commercial compounds was docked to
the ubiquitin binding pocket of HDAC6 with Glide20
■
(Schrodinger, NY). Two virtual screens were conducted in
̈
B
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Table 1. FP IC50 and Ligand Efficiency (LE) for Exemplar Compounds
a
b
Source of the virtual hits: + = original selection; ++ = follow-up of the original selection. Ligand efficiency (in kcal mol−1 atom−1) was calculated as
c
LE = (1.37 × pIC50)/HA. The logD was calculated using ChemAxon’s JChem for Excel, version 16.11.2800.3575; IC50 determination experiments
were performed in triplicate, and the values are presented as mean SD; NB = no binding; HA = heavy atom.
acid in 7 was not tolerated. The sulfur of compound 2 could be
replaced with a hydrophobic N-methyl group in 8. However, its
analogue (9) without the methyl group as well as other isosteres
of the benzothiazole ring (10 and 11) were inactive. Compounds
12−18 exemplify other active scaffolds, with IC50’s ranging from
44 to 320 μM. Finally, the unrelated hits 19−20 were weak
inhibitors of HDAC6 ZnF-UBD.
The binding of 1 was further confirmed by surface plasmon
resonance (SPR) with a KD of 8 μM. We also obtained a crystal
structure of 1 in complex with the ZnF-UBD of HDAC6.
Although 1 may be modeled trivially in the Fo − Fc map,
additional density was observed on the thiazole ring distal to the
ubiquitin binding pocket, which was attributed to possible
radiation damage of the ligand as described in the PDB
annotation of this structure.
As shown in Figure 2a, the carboxylate of 1 overlaps well with
the one from the ubiquitin peptide and forms direct or water-
mediated hydrogen bonds with R1155, Y1184, L1162, and a salt
bridge with R1155. Additionally, the aromatic scaffold of the
compound makes π-stacking interactions with W1182 and
R1155. Upon binding of 1, R1155 and Y1156 adopted a
conformation that is closer to the apo state than to the ubiquitin-
bound state. The main difference is that the guanidine group of
R1155 is rotated 90° to form a cation-π interaction with the
C
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 2. Crystal structure of HDAC6 ZnF-UBD in complex with (a) 1 (PDB: 6CE6) and (b) 2 (PDB: 6CEF). The binding pocket residues (gray) and
the inhibitors (cyan) are displayed as ball sticks and the hydrogen bonds as green dashed lines. The C-terminal glycine of the ubiquitin peptide (PDB:
3GV4) is shown as magenta sticks.
Figure 3. Design strategy for the hit expansion: (a) Substructure search and ring expansion of analogues of 2; (b) scaffold-hopping followed by
substructure search.
aromatic ring and a hydrogen bond with the buried carboxylic
acid of the ligand (Figure S1). It should be noted that the
carboxylate moiety of this ligand distal to the ubiquitin binding
pocket, although not fully resolved in the electron density, may
form interactions with the adjacent molecule of HDAC6 in the
crystal lattice which could affect its binding mode as resolved in
the crystal structure. However, the close analogue 2, which lacks
this carboxylate, binds in the same orientation as 1, but was 36-
fold weaker, indicating that R1155 can productively interact with
the exposed carboxylate (Figure 2a).
which is a substructure of 1, and although weaker (IC50 = 180
52 μM), has a slightly better ligand efficiency. We solved the
crystal structure of 2 in complex with HDAC6 ZnF-UBD (Figure
2b). As expected, fragment 2 adopted a similar binding pose and
recapitulated favorable interactions observed with 1.
To circumvent the lack of close commercial analogues of 2,
several compound design strategies were employed to identify
structurally related molecules of potential interest. First, we
performed a substructure search focused on 2 (Figure 3a), which
resulted in 111 compounds. Supported by our SAR, we envisaged
that a ring expansion from the five-member ring to a six-member
ring in 2 could also lead to attractive molecules. This
substructure-search yielded 423 compounds (Figure 3a). Both
substructure searches were designed to retrieve compounds with
one or two carbon atoms connecting the carboxylate group and
the heterocycle, as our SAR showed that compounds differing
only in the linker size had similar IC50 values (for example: 12 vs
HIT EXPANSION
■
Although compound 1 was the most potent inhibitor, it was not a
favorable starting point for developing a chemical probe, as the
two acidic moieties make the compound highly hydrophilic with
a calculated (ChemAxon) logD and polar surface area (PSA) of
−4.0 and 106 Å2, respectively. We focused instead on fragment 2,
D
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Table 2. Follow-ups of 2 Are Shown, and Annotated with IC50 Values That Were Determined Using a Fluorescence Polarization
Assay
18, Table 1). We also used the scaffold-hopping tool BROOD21
(OpenEye, CA) to identify additional chemical starting points
(Figure 3b). Not surprisingly, a sulfonyl group emerged as a
possible isostere of the carboxylate (Figure S2), which was an
attractive mechanism to improve the physicochemical properties
of the compound, at the risk of losing electrostatic interactions
with R1155. A substructure search for analogues of 2 with the
sulfonyl group yielded 315 commercial compounds (Figure 3b).
The resulting library of 849 compounds was then docked to
the structure of HDAC6 ZnF-UBD in complex with 1 using
Glide20 (Schrodinger, NY), with three hydrogen bond
constraints at the side chain OH of Y1184, the NH backbone
and the side chain of R1155. After careful visual inspection, 100
compounds that satisfied the hydrogen bond constraints and also
formed a π-stacking interaction with W1182 and R1155 were
selected and rescored using the molecular mechanics generalized
Born surface area (MM/GBSA) method with AMBER 12.22
Upon visual inspection, 13 compounds were ordered, and tested
using a fluorescence polarization peptide displacement assay. Of
the 13 compounds ordered, 9 (21, 22, 24, 26−28, 31−33)
displaced the ubiquitin peptide at micromolar concentration
(Table 2).
The data provided further insight into the structure−activity
relationships of this chemical series. Adding a chlorine at position
C5 of the benzothiazole ring resulted in a 3-fold gain in potency
(compound 21), but removing a methylene from the acidic side
chain was detrimental (compound 22). Compound 24, with a 3-
methoxyquinoxaline scaffold, inhibited HDAC6 ZnF-UBD with
an IC50 of 45 μM. Comparing 2 vs 22, and 23 vs 25 indicated that
a propionic was preferred over an acetic acid. The docking poses
E
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 4. Crystal structures of four fragments in complex with HDAC6 ZnF-UBD: 21 (PDB: 5KH3); 22 (PDB: 6CE8); 23 (PDB: 6CEA); 24 (PDB:
6CEC). The compounds (cyan) and the amino acid residues (gray) are displayed as ball and sticks. Hydrogen bonds are displayed as green dashed lines.
The lower occupancy of R1155 in 22 is displayed as green ball and sticks.
Figure 5. Crystal structures of 31 (PDB: 6CED) and 33 (PDB: 6CEE) in complex with HDAC6 ZnF-UBD. The compounds (cyan) and the amino acid
residues (gray) are displayed as ball and sticks. Hydrogen bonds are displayed as green dashed lines. The lower occupancy of R1155 in 31 is displayed as
green ball and sticks.
F
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Table 3. Structure, IC50, and LE Values of Compounds Used To Investigate the Importance of the Fused Ring and the Carboxylate
Group
Figure 6. Crystal structure of HDAC6 ZnF-UBD in complex with 24 (PDB: 6CEC) (a) and 28 (PDB: 5WBN) (b, c), respectively. The compound
(cyan) and the amino acid residues (gray) are displayed as ball and sticks. Hydrogen bonds are displayed as green dashed lines. The surface is colored by
binding property: white = neutral; green = hydrophobic; blue = hydrogen bonding donor potential; red = hydrogen bond acceptor potential.
of compounds 26−28 showed that they could extend beyond the
ubiquitin C-terminus binding pocket, but only 28 inhibited the
protein with an IC50 of 25 μM (LE = 0.23). Finally, the two
sulfonyl compounds (29 and 30) were inactive.
The quinazolinone scaffold of 31 represented a significant
breakthrough in the potency and ligand efficiency of this
chemical series with an IC50 of 2.3 μM by FP (LE = 0.45), and a
KD of 1.5 μM determined by SPR. Removing the critical N-
methyl group at the amide moiety (compound 32) or shifting its
position as in the regioisomer 33, resulted in a 100- or 28-fold
loss in potency, respectively.
To investigate the structural basis for the relative potencies, all
13 compounds were soaked into crystals of HDAC6 ZnF-UBD,
and 7 crystal structures (compounds 21−24, 28, 31, and 33)
were obtained with resolutions ranging from 1.55 to 1.75 Å. The
docked poses were confirmed for six of the seven structures
(RMSD 0.3−1.9 Å; Figure S3) and recapitulated the general
binding mode previously observed with fragment 2, including an
anchored carboxylate warhead, salt bridges, and a cation-π
interaction with R1155, as well as a π-stacking interaction with
W1182 (Figures 4 and 5).
the benzothiazole rings of the two compounds overlapped
perfectly (Figure 4). The side chain of R1155 was also found in
an alternate, lower-occupancy conformation in complex with 22,
where the guanidine group was flipped ∼180°. On the basis of
the crystal structures, the increased potency of 2 vs 22 probably
reflects a reduced conformational strain associated with a longer
carboxylic chain.
The amino acid side-chains lining the inhibitor binding pocket
of HDAC6 ZnF-UBD in complex with 31 adopted conforma-
tions identical to those seen in previous structures. Also, as seen
with compound 22, electron density was observed for two
partial-occupancy conformations of R1155 (Figure 5). The
critical methyl group of 31 occupies a cavity formed by the polar
guanidine of R1155 and the hydroxyl group of Y1189, as well as
the partially hydrophobic side-chain of W1182. The combined
loss of hydrophobic interactions with W1182 and electronic
rearrangement at the aromatic system were probably responsible
for the 100-fold decrease in potency upon removal of this methyl
group (compound 32). The gain of a hydrogen bond with the
carbonyl group of the regioisomer 33 (Figure 5) partially
mitigates this loss (28-fold drop in potency).
As expected, 21 binds in the same conformation as 2; the
chlorine group pointed toward the solvent and is probably
extending or strengthening π−π interactions with neighboring
side-chains (Figure 4). Fragment 22, featuring an acetic instead
of a propionic acid, preserved the interactions observed for 2 and
A BICYCLIC RING SYSTEM AND A CARBOXYLATE
ARE ESSENTIAL FOR BINDING
■
The complex of HDAC6 ZnF-UBD with 24 was also solved: the
methoxy group of the fragment points toward the side chain of
G
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Y1189, but does not form productive interactions with this
residue (Figure 4). Only one of the two aromatic rings of 24
makes π-stacking and cation-π interactions with W1182 and
R1155, respectively, while both rings were well positioned to
make these interactions in the weaker compound 23 (Figure 4).
This raised the possibility that these fragments could be
simplified to a monocyclic ring system. To test this hypothesis
we synthesized compound 34, the single-ring analogue of 24
(Table 3). No detectable binding was observed, supporting the
notion that a fused bicyclic aromatic ring system is necessary.
Replacement of the carboxylate with a primary or secondary
amide (35 and 36, respectively) or an ester group (37) was not
tolerated, indicating that the conserved network of direct and
water-mediated interactions between the carboxylate and R1155,
Y1184, and L1162 is essential for binding (Figure 4−5, Table 3).
Figure 7. FP competition assay using increasing concentrations of
compound 31, a FITC-labeled RLRGG peptide (50 nM) and full-length
HDAC6 (1 μM). An IC50 of 3.5 0.9 μM was obtained from the average
of three independent measurements.
CONFORMATIONAL REARRANGEMENTS CAN
OPEN THE DOOR TO A SIDE POCKET
■
side-pocket that may be exploited to increase potency. We
believe that the chemical inhibitors presented here and
accompanying crystal structures will inform the discovery of a
novel class of HDAC6-selective chemical probes.
The crystal structures of HDAC6 ZnF-UBD feature a well-
defined cavity juxtaposed to the binding pocket exploited by the
compounds presented here. This secondary pocket is separated
from the primary pocket by a wall composed of Y1189 and
R1155 (Figure 6a). Interestingly, in the structure in complex with
28, this wall disappears, raising the possibility of simultaneously
exploiting both cavities (Figure 6b). Here, R1155 adopts an
unique conformation (Figure S4) where it flips away from the
secondary pocket to stack with a third fused ring of the central
core (absent from other inhibitors) of 28, opening the door
between the two binding pockets (Figure 6c).
We have recently reported weak fragments occupying the
secondary pocket, and a weak inhibitor was able to bind both
pockets suggesting that it is ligand-able.17 However, the stacking
interaction of the heterocyclic core with the R1155 and W1182
was disrupted in this inhibitor resulting in weak affinity.
Compound 28 is not a favorable scaffold due to chemical
liabilities, but the crystal structure and associated IC50 of 25 μM
demonstrates the possibility to grow the fragments identified in
this work into the side-pocket without disrupting the key
interactions in the main pocket, which could be a promising
strategy to improve potency.
EXPERIMENTAL SECTION
■
Primary Virtual Screening. Ligprep23 (Schrodinger, NY) was used
to convert to 3D a diverse virtual library of 1 280 604 drug-like
compounds from commercial vendors (protonation states at pH 7.3
0.2 were generated with Epik). The resulting library was docked to the
ubiquitin-binding pocket of HDAC6 (PDB code 3GV4) with Glide
HTVS and the top 10% scoring compounds were redocked with Glide
SP (Schrodinger, NY).20 The process was conducted twice
independently, with or without a water molecule at the bottom of the
binding pocket (forming a hydrogen-bond with the main-chain of
C1153 and L1162). The 213 compounds scoring better than −7.0 in the
presence of the water molecule and the 100 compounds scoring better
than −6.5 in the absence of water were selected for visual inspection,
from which 40 compounds were subjectively picked, based on chemical
scaffold and binding pose, and 33 were in stock and ordered.
Substructure Search. FILTER (OpenEye, CA)24 was used to
perform substructure searches using the SELECT keyword, which
specifies a substructure definition in the form of a SMARTS string to be
present in all molecules that pass the filter. The substructure searches
were run against ∼22 million commercial compounds downloaded from
the ZINC database.25 LigPrep23 was used to prepare the ligands using
default settings.
Scaffold Hopping. BROOD21 (OpenEye, CA) was used to select
potential isosteric groups of the carboxylic acid using the fragment
[R1]CCC(O)[O−] as the query. BROOD generates analogues of a
compound by replacing selected fragments (scaffolds) in the molecule
with fragments that have similar shape and electrostatics. The sulfonyl
group was the best group to match the query fragment in terms of shape,
electrostatics, and attachment geometry.
Ligand Docking. The library designed by the substructure search
and scaffold hopping exercise was docked to the HDAC6 ZnF-UBD
structure in complex with 1 using Glide with default settings. Three
hydrogen bond constraints at the side chain OH of Y1184, the NH
backbone, and the side chain of R1155 were used in the docking
calculations.
Molecular Mechanics (MM/GBSA). The docked poses of the
ligands in complex with HDAC6 ZnF-UBD were rescored using the
MM/GBSA method.26 Missing atoms and hydrogens were added using
the tleap program in Amber12 software.22 Amber ff99SB27 force field
and GAFF force field28,29 were assigned to amino acid residues and to
the ligands, respectively. The GAFF parameters were generated using
antechamber and parmchk program, and partial atomic charges were
derived using AM1-BCC methodology.30 Each complex structure was
solvated in a TIP3P water cubic box extending at least 10 Å from the
complex, and then neutralized by the addition of appropriate number of
monovalent counterions.
COMPOUND 31 DISPLACES THE RLRGG PEPTIDE
FROM FULL-LENGTH HDAC6
■
We also used a fluorescence polarization assay to test whether
fragment 31 would bind to the ZnF-UBD domain in the context
of full-length HDAC6 (HDAC61−1215), a prerequisite for the
future development of a cell-active chemical probe. Our result
showed that 31 also displaces the C-terminal ubiquitin peptide
from full length HDAC6 with an IC50 of 3.5 0.9 μM (Figure 7).
This value showed no significant difference from the 2.3 0.6
μM obtained with the ZnF-UBD domain alone
(HDAC61109−1215) and shows that 31 does bind to HDAC6 in
the context of the full-length protein.
CONCLUSION
■
Here, we report a successful attempt to target the ZnF-UBD of
HDAC6 with small molecules selected by virtual screening. SAR
analysis indicates that a carboxylate warhead, linked to a fused
bicyclic aromatic ring system that can stack with neighboring
tryptophan and arginine side-chains are important chemical
features for activity. Inhibitor-induced conformational remodel-
ing at the binding site can open-up a channel toward a ligand-able
H
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Cloning, Protein Expression, and Purification. DNAs encoding
HDAC61109−1213 and HDAC61109−1215 were subcloned into a modified
pET28 vector encoding a thrombin cleavable (GenBank EF442785) N-
terminal His6-tag (pET28-LIC) and HDAC61109−1215 was also
subcloned into a modified pET28 vector encoding an N-terminal
AviTag for in vivo biotinylation and a C-terminal His6-tag (p28BIOH-
LIC) using a ligation-independent InFusion cloning kit (ClonTech) and
verified by DNA sequencing. Proteins were overexpressed in BL21
(DE3) Codon Plus RIL Escherichia coli (Agilent). Cultures were grown
in M9 minimal media supplemented with 50 μM ZnSO4 and also 10 μg/
mL biotin for HDAC61109−1215 p28BIOH-LIC expression. Expression
cultures were induced using 0.5 mM IPTG overnight at 15 °C. Proteins
were purified using nickel-nitrilotriacetic acid (Ni-NTA) agarose resin
(Qiagen), and the tag was removed by thrombin for HDAC61109−1213
pET28-LIC. Uncleaved proteins and thrombin were removed by
another pass with Ni-NTA resin. Proteins were further purified using gel
filtration (Superdex 75, GE Healthcare). The final concentrations of
purified proteins were 5−10 mg/mL as measured by UV absorbance at
280 nm.
min before FP analysis. Previous assay optimization with titration of
protein concentration at 50 nM peptide showed these conditions gave
∼85% maximum FP.
RLRGG Peptide and Full-Length HDAC6 Assay. Experiments were
performed in a total volume of 10 μL in a 384-well black polypropylene
PCR plate (Axygen). FP was measured using Synergy 4 (BioTek) after
10 min of incubation. The excitation and emission wavelengths were 485
and 528 nm, respectively. Each well contained 9 μL of compound 31
(200 μM) in buffer containing 20 mM Hepes pH 7.4, 150 mM NaCl,
0.005% (v/v) Tween-20, 0.1% (v/v) DMSO serially diluted. 1 μL of 10
μM HDAC61−1215 and 500 nM N-terminally FITC-labeled RLRGG was
then added to each well. The plate was centrifuged at 150g for 1 min
before FP analysis.
Chemistry and Compound Purity. Compounds 21, 24, and 31−
36 were synthesized according to the procedures described in the
Supporting Information. All commercial and synthesized compounds
tested in vitro were ≥95% pure. Purity was determined by analytical
HPLC on a Agilent 1100 series instrument equipped with a
Phenomenex KINETEX column (50.0 mm × 4.6 mm, C18, 2.6 μM)
at 25 °C. A linear gradient starting from 5% acetonitrile and 95% water
to 95% acetonitrile and 5% water over 4 min followed by elution at 95%
acetonitrile and 5% water was employed. Formic acid (0.1%) was added
to all solvents.
DNA encoding HDAC61−1215 was cloned into a modified derivative
pFastBac Dual vector (Invitrogen) encoding an N-terminal AviTag and
a C-terminal His6 tag (pFBD-BirA). Cloning was completed using a
ligation-independent InFusion cloning kit (ClonTech) and verified by
DNA sequencing. HDAC61−1215 was overexpressed in sf9 insect cells.
Cultures were grown in HyQ SFX Insect Serum Free Medium (Fisher
Scientific) to a density of 4 × 106 cells/mL and infected with 10 mL of P3
viral stock media per 1 L of cell culture. Cell culture medium was
collected after 4 days of incubation in a shaker. Protein was purified
using nickel-nitrilotriacetic acid (Ni-NTA) agarose resin (Qiagen).
Following overnight dialysis, protein was further purified using anion
exchange (HiTrap Q HP, GE Healthcare) and gel filtration (Superdex
75, GE Healthcare). The final concentration of purified HDAC61−1215
was 3.5 mg/mL as measured by UV absorbance at 280 nm.
ASSOCIATED CONTENT
■
S
* Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.jmed-
chem.8b00258.
Additional data regarding supplementary tables and
figures. Data collection and refinement for compounds
1, 2, 22−24, 28, 31, and 33 (Table S1). General practical
considerations, synthetic procedures, characterization
Crystallization. The apo crystal form of HDAC61109−1215 (PDB ID:
3C5K) was previously reported.18 These crystals can be used to seed for
the crystal form amenable to soaking. Diluting 1 μL crystal suspension
1:10000 with mother liquor and vortexing the sample vigorously yielded
a seed mix. HDAC61109−1213 can be crystallized in 2 M Na formate, 0.1 M
Na acetate pH 4.6, 5% ethylene glycol in 5:4:1 3.5 mg/mL protein,
mother liquor, and seed mix per drop. HDAC61109−1213 crystals were
soaked by adding 5% (v/v) of a 200 mM or 400 mM DMSO-solubilized
stock of compounds to the drop for 2 h prior to mounting and cryo-
cooling.
1
data, H NMR, 13CNMR, and HPLC purity analysis for
the compounds 21, 24, 31−36 (PDF)
Molecular formula strings and the associated biological
data (CSV)
Crystallographic information files (ZIP)
Accession Codes
Atomic coordinates and experimental data were deposited in the
Protein Databank and will be released upon article publication.
HDAC6 ZnF-UBD cocrystal structure PDB codes: 1 (6CE6), 2
(6CEF), 22 (6CE8), 23 (6CEA), 24 (6CEC), 28 (5WBN), 31
(6CED), and 33 (6CEE).
Data Collection, Structure Determination, and Refinement.
X-ray diffraction data for HDAC61109−1213 cocrystals were collected at
100 K at Rigaku FR-E Superbright home source at a wavelength of
1.54178 Å. All data sets were processed with XDS31 and Aimless.32
Models were refined with cycles of COOT33 for model building and
visualization, with REFMAC,34 for restrained refinement and validated
with MOLPROBITY.35
Surface Plasmon Resonance (SPR). Studies were performed using
a Biacore T200 (GE Health Sciences). Approximately 5000 response
units (RU) of biotinylated HDAC61109−1215 were coupled onto one flow
cell of a SA chip as per manufacturer’s protocol, and an empty flow cell
used for reference subtraction. Two-fold serial dilutions of compounds
were prepared in 10 mM HEPES pH 7.4, 150 mM NaCl, 0.005% (v/v)
Tween-20, 1% (v/v) DMSO. KD determination experiments were
performed using single-cycle kinetics with 30 s contact time, 30 μL/min
flow rate at 20 °C. KD values were calculated using steady state affinity
fitting and the Biacore T200 Evaluation software.
Fluorescence Polarization (FP) Displacement Assay. RLRGG
Peptide and HDAC6 ZnF-UBD. All experiments were performed in 384-
well black polypropylene PCR plates (Axygen) in 10 μL volume.
Fluorescence polarization (FP) was measured using a BioTek Synergy 4
(BioTek), and excitation and emission wavelengths were 485 and 528
nm, respectively. In each well, 9 μL compound solutions in buffer
containing 10 mM HEPES pH 7.4, 150 mM NaCl, 1% (v/v) DMSO
were serially diluted. 1 μL of 30 μM HDAC61109−1215 and 500 nM N-
terminally FITC-labeled RLRGG were then added to each well.
Following 1 min centrifugation at 250g, the assay was incubated for 10
AUTHOR INFORMATION
■
Corresponding Author
*Phone: +1 416-978-3092. E-mail: matthieu.schapira@
utoronto.ca.
ORCID
Renato Ferreira de Freitas: 0000-0001-6195-4236
Rachel J. Harding: 0000-0002-1134-391X
Ivan Franzoni: 0000-0001-8110-6218
Mark Lautens: 0000-0002-0179-2914
Matthieu Schapira: 0000-0002-1047-3309
Author Contributions
#R.F.d.F. and R.J.H. contributed equally to this work. R.F.F., V.S.,
and M.S. designed compounds, R.F.F. and M.S. wrote the
manuscript, R.H. and M.R. solved structures, R.H., H.U., and
M.M. tested compounds, and I.F. and M.L. synthesized
compounds. C.H.A. advised throughout the project. All authors
have given approval to the final version of the manuscript.
I
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
and Cell Viability in Response to Misfolded Protein Stress. Cell 2003,
115, 727−738.
(12) Li, Y.; Shin, D.; Kwon, S. H. Histone Deacetylase 6 Plays a Role as
a Distinct Regulator of Diverse Cellular Processes. FEBS J. 2013, 280,
775−793.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
(13) Santo, L.; Hideshima, T.; Kung, A. L.; Tseng, J.-C.; Tamang, D.;
Yang, M.; Jarpe, M.; van Duzer, J. H.; Mazitschek, R.; Ogier, W. C.;
Cirstea, D.; Rodig, S.; Eda, H.; Scullen, T.; Canavese, M.; Bradner, J.;
Anderson, K. C.; Jones, S. S.; Raje, N. Preclinical Activity,
Pharmacodynamic, and Pharmacokinetic Properties of a Selective
HDAC6 Inhibitor, ACY-1215, in Combination with Bortezomib in
Multiple Myeloma. Blood 2012, 119, 2579−2589.
The SGC is a registered charity (Number 1097737) that receives
funds from AbbVie, Bayer Pharma AG, Boehringer Ingelheim,
Canada Foundation for Innovation, Eshelman Institute for
Innovation, Genome Canada through Ontario Genomics
Institute [OGI-055], Innovative Medicines Initiative (EU/
EFPIA) [ULTRA-DD Grant No. 115766], Janssen, Merck
KGaA, Darmstadt, Germany, MSD, Novartis Pharma AG,
Ontario Ministry of Research, Innovation and Science (MRIS),
(14) Hideshima, T.; Qi, J.; Paranal, R. M.; Tang, W.; Greenberg, E.;
West, N.; Colling, M. E.; Estiu, G.; Mazitschek, R.; Perry, J. A.; Ohguchi,
Pfizer, Sao Paulo Research Foundation-FAPESP, Takeda, and
̃
H.; Cottini, F.; Mimura, N.; Gorgun, G.; Tai, Y.-T.; Richardson, P. G.;
̈
̈
Wellcome [106169/ZZ14/Z]. I.F. and M.M. are supported by
NSERC CREATE Grant 432008-2013. Andrei Yudin (Uni-
versity of Toronto) is thanked for analytical HPLC access. We
are grateful to OpenEye Scientific Software, Inc. for providing us
with an academic license for their software.
Carrasco, R. D.; Wiest, O.; Schreiber, S. L.; Anderson, K. C.; Bradner, J.
E. Discovery of Selective Small-Molecule HDAC6 Inhibitor for
Overcoming Proteasome Inhibitor Resistance in Multiple Myeloma.
Proc. Natl. Acad. Sci. U. S. A. 2016, 113, 13162−13167.
(15) Amengual, J. E.; Johannet, P.; Lombardo, M.; Zullo, K.; Hoehn,
D.; Bhagat, G.; Scotto, L.; Jirau-Serrano, X.; Radeski, D.; Heinen, J.;
Jiang, H.; Cremers, S.; Zhang, Y.; Jones, S.; O’Connor, O. A. Dual
Targeting of Protein Degradation Pathways with the Selective HDAC6
Inhibitor ACY-1215 and Bortezomib Is Synergistic in Lymphoma. Clin.
Cancer Res. 2015, 21, 4663−4675.
(16) Senger, J.; Melesina, J.; Marek, M.; Romier, C.; Oehme, I.; Witt,
O.; Sippl, W.; Jung, M. Synthesis and Biological Investigation of Oxazole
Hydroxamates as Highly Selective Histone Deacetylase 6 (HDAC6)
Inhibitors. J. Med. Chem. 2016, 59, 1545−1555.
(17) Harding, R. J.; Ferreira de Freitas, R.; Collins, P.; Franzoni, I.;
Ravichandran, M.; Ouyang, H.; Juarez-Ornelas, K. A.; Lautens, M.;
Schapira, M.; von Delft, F.; Santhakumar, V.; Arrowsmith, C. H. Small
Molecule Antagonists of the Interaction between the Histone
Deacetylase 6 Zinc-Finger Domain and Ubiquitin. J. Med. Chem.
2017, 60, 9090−9096.
ABBREVIATIONS USED
■
FP, fluorescence polarization; KD, dissociation constant; SPR,
surface plasmon resonance; MM/GBSA, molecular mechanics
generalized Born surface area; GAFF, general AMBER force field
REFERENCES
■
(1) Miyake, Y.; Keusch, J. J.; Wang, L.; Saito, M.; Hess, D.; Wang, X.;
Melancon, B. J.; Helquist, P.; Gut, H.; Matthias, P. Structural Insights
into HDAC6 Tubulin Deacetylation and Its Selective Inhibition. Nat.
Chem. Biol. 2016, 12, 748−754.
(2) Haberland, M.; Montgomery, R. L.; Olson, E. N. The Many Roles
of Histone Deacetylases in Development and Physiology: Implications
for Disease and Therapy. Nat. Rev. Genet. 2009, 10, 32−42.
(3) Hai, Y.; Christianson, D. W. Histone Deacetylase 6 Structure and
Molecular Basis of Catalysis and Inhibition. Nat. Chem. Biol. 2016, 12,
741−747.
(4) Hubbert, C.; Guardiola, A.; Shao, R.; Kawaguchi, Y.; Ito, A.; Nixon,
A.; Yoshida, M.; Wang, X.-F.; Yao, T.-P. HDAC6 Is a Microtubule-
Associated Deacetylase. Nature 2002, 417, 455−458.
(5) Kovacs, J. J.; Murphy, P. J. M.; Gaillard, S.; Zhao, X.; Wu, J.-T.;
Nicchitta, C. V.; Yoshida, M.; Toft, D. O.; Pratt, W. B.; Yao, T.-P.
HDAC6 Regulates Hsp90 Acetylation and Chaperone-Dependent
Activation of Glucocorticoid Receptor. Mol. Cell 2005, 18, 601−607.
(6) Zhang, X.; Yuan, Z.; Zhang, Y.; Yong, S.; Salas-Burgos, A.; Koomen,
J.; Olashaw, N.; Parsons, J. T.; Yang, X.-J.; Dent, S. R.; Yao, T.-P.; Lane,
W. S.; Seto, E. HDAC6 Modulates Cell Motility by Altering the
Acetylation Level of Cortactin. Mol. Cell 2007, 27, 197−213.
(7) Choi, H.; Kim, H. J.; Kim, J.; Kim, S.; Yang, J.; Lee, W.; Park, Y.;
Hyeon, S. J.; Lee, D. S.; Ryu, H.; Chung, J.; Mook-Jung, I. Increased
Acetylation of Peroxiredoxin1 by HDAC6 Inhibition Leads to Recovery
of Aβ-Induced Impaired Axonal Transport. Mol. Neurodegener. 2017, 12,
23.
(18) Ouyang, H.; Ali, Y. O.; Ravichandran, M.; Dong, A.; Qiu, W.;
MacKenzie, F.; Dhe-Paganon, S.; Arrowsmith, C. H.; Zhai, R. G. Protein
Aggregates Are Recruited to Aggresome by Histone Deacetylase 6 via
Unanchored Ubiquitin C Termini. J. Biol. Chem. 2012, 287, 2317−2327.
(19) Ferreira de Freitas, R.; Schapira, M. A Systematic Analysis of
Atomic Protein−ligand Interactions in the PDB. MedChemComm 2017,
8, 1970−1981.
(20) Glide, version 6.7; Schrodinger, LLC: New York, 2015.
̈
(21) BROOD; OpenEye Scientific Software: Santa Fe, NM, 2015.
(22) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C.
L.; Wang, J.; Duke, R. E.; Luo, R.; Walker, R. C.; Zhang, W.; Merz, K. M.;
Robertson, B.; Wang, B.; Hayik, S.; Roitberg, A.; Seabra, G.; Kolossvary,
I.; Wong, K. F.; Paesani, F.; Vanicek, J.; Wolf, R. M.; Liu, J.; Wu, X.;
Brozell, S.; Steinbrecher, T.; Gohlke, H.; Cai, Q.; Ye, X.; Wang, J.; Hsieh,
M.-J.; Cui, G.; Roe, D. R.; Mathews, D. H.; Seetin, M. G.; Salomon-
Ferrer, R.; Sangui, C.; Babin, V.; Luchko, T.; Gusarov, S.; Kovalenko, A.;
Kollman, P. A. AMBER 12, University of California: San Francisco,
2012.
(23) LigPrep, version 3.5; Schrodinger, LLC: New York, 2015.
̈
(24) Filter, 2.4; OpenEye Scientific Software, Santa Fe, NM.
(25) Irwin, J. J.; Shoichet, B. K. ZINC - A Free Database of
Commercially Available Compounds for Virtual Screening. J. Chem. Inf.
Model. 2005, 45, 177−182.
(8) Cohen, T. J.; Guo, J. L.; Hurtado, D. E.; Kwong, L. K.; Mills, I. P.;
Trojanowski, J. Q.; Lee, V. M. Y. The Acetylation of Tau Inhibits Its
Function and Promotes Pathological Tau Aggregation. Nat. Commun.
2011, 2, 252.
(26) Ferreira de Freitas, R.; Eram, M. S.; Smil, D.; Szewczyk, M. M.;
Kennedy, S.; Brown, P. J.; Santhakumar, V.; Barsyte-Lovejoy, D.;
Arrowsmith, C. H.; Vedadi, M.; Schapira, M. Discovery of a Potent and
Selective Coactivator Associated Arginine Methyltransferase 1
(CARM1) Inhibitor by Virtual Screening. J. Med. Chem. 2016, 59,
6838−6847.
(27) Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.;
Simmerling, C. Comparison of Multiple Amber Force Fields and
Development of Improved Protein Backbone Parameters. Proteins:
Struct., Funct., Genet. 2006, 65, 712−725.
(9) Serrador, J. M.; Cabrero, J. R.; Sancho, D.; Mittelbrunn, M.;
́
Urzainqui, A.; Sanchez-Madrid, F. HDAC6 Deacetylase Activity Links
the Tubulin Cytoskeleton with Immune Synapse Organization.
Immunity 2004, 20, 417−428.
(10) Banerjee, I.; Miyake, Y.; Nobs, S. P.; Schneider, C.; Horvath, P.;
Kopf, M.; Matthias, P.; Helenius, A.; Yamauchi, Y. Influenza A Virus
Uses the Aggresome Processing Machinery for Host Cell Entry. Science
2014, 346, 473−477.
(11) Kawaguchi, Y.; Kovacs, J. J.; McLaurin, A.; Vance, J. M.; Ito, A.;
Yao, T. P. The Deacetylase HDAC6 Regulates Aggresome Formation
J
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
(28) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A.
Development and Testing of a General Amber Force Field. J. Comput.
Chem. 2004, 25, 1157−1174.
(29) Wang, J.; Wang, W.; Kollman, P. A.; Case, D. A. Automatic Atom
Type and Bond Type Perception in Molecular Mechanical Calculations.
J. Mol. Graphics Modell. 2006, 25, 247−260.
(30) Jakalian, A.; Bush, B. L.; Jack, D. B.; Bayly, C. I. Fast, Efficient
Generation of High-Quality Atomic Charges. AM1-BCC Model: I.
Method. J. Comput. Chem. 2000, 21, 132−146.
(31) Kabsch, W. XDS. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2010,
66, 125−132.
(32) Evans, P. R.; Murshudov, G. N. How Good Are My Data and
What Is the Resolution? Acta Crystallogr., Sect. D: Biol. Crystallogr. 2013,
69, 1204−1214.
(33) Emsley, P.; Lohkamp, B.; Scott, W. G.; Cowtan, K. Features and
Development of Coot. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2010,
66, 486−501.
́
(34) Murshudov, G. N.; Skubak, P.; Lebedev, A. A.; Pannu, N. S.;
Steiner, R. A.; Nicholls, R. A.; Winn, M. D.; Long, F.; Vagin, A. A.
REFMAC5 for the Refinement of Macromolecular Crystal Structures.
Acta Crystallogr., Sect. D: Biol. Crystallogr. 2011, 67, 355−367.
(35) Chen, V. B.; Arendall, W. B.; Headd, J. J.; Keedy, D. A.;
Immormino, R. M.; Kapral, G. J.; Murray, L. W.; Richardson, J. S.;
Richardson, D. C. MolProbity: All-Atom Structure Validation for
Macromolecular Crystallography. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 2010, 66, 12−21.
K
DOI: 10.1021/acs.jmedchem.8b00258
J. Med. Chem. XXXX, XXX, XXX−XXX