Role of Fluorination in the Histone Deacetylase 6 (HDAC6) Selectivity of Benzohydroxamate-Based Inhibitors

Abstract

Nonselective histone deacetylase (HDAC) inhibitors show dose-limiting side effects due to the inhibition of multiple, essential HDAC subtypes that can be limited or prevented by restricting their selectivity. We herein report the crystal structures of zebrafish HDAC6 catalytic domain 2 (zHDAC6-CD2) in complex with the selective HDAC6 inhibitors ITF3756 and ITF3985 and shed light on the role of fluorination in the selectivity of benzohydroxamate-based structures over class I isoforms. The reason for the enhancement in the selectivity of the benzohydroxamate-based compounds is the presence of specific interactions between the fluorinated linker and the key residues Gly582, Ser531, and His614 of zHDAC6, which are hindered in class I HDAC isoforms by the presence of an Aspartate that replaces Ser531. These results can be used in the design and development of novel, highly selective HDAC6 inhibitors.

Histone deacetylases (HDACs) are a family of enzymes that were originally discovered because of their role in the hydrolysis of acetyl-l-lysine side chains in the N-terminal region of core histones. Mammals express 18 different HDAC enzymes that can be grouped into two mechanistically distinct families: the 7 NAD-dependent members of the sirtuin family and the 11 Zn-dependent HDACs, which are conventionally partitioned into class I (HDAC1, 2, 3, and 8), class IIa (HDAC4, 5, 7, and 9), class IIb (HDAC6 and 10), and class IV (HDAC11).¹

HDACs are found in different cellular compartments and have been found to deacetylate, or more in general deacylate, different histone or nonhistone protein or nonprotein substrates.¹ The acetylation status of histones is involved in the regulation of chromatin accessibility to transcription factor complexes and therefore gene expression,² whereas acetylation of nonhistone proteins affects their function and stability and may strongly impact cell physiology.³ HDAC inhibitors (HDACis) are considered promising therapeutic alternatives to the available cancer chemotherapies, and several inhibitors of Zn-dependent HDACs have been approved for the treatment of human malignancies.⁴ The first generation of compounds inhibit all or most of the HDAC isoforms and cause side effects including hematological and gastrointestinal tract toxicity and fatigue, that limit their use in many indications. These unwanted effects have recently driven investigations toward subtype-selective molecules. In this regard, HDAC6 represents an attractive target since HDAC6 knockout (KO) mice develop normally and have no signs of pathological alterations.⁵ Furthermore, HDAC6 inhibitors have been reported to be well tolerated, suggesting a wider therapeutic window compared to other HDACis.⁶ HDAC6 activity has been shown to be associated with several pathological conditions spanning from neuropathies to immune modulation, respiratory dysfunction, and cancer.⁵

HDAC6 has two catalytic domains (CDs):⁷ one, conventionally indicated as CD1, has been proposed to accept C-terminally acetylated lysines, whereas the other, named CD2, has been reported to deacetylate α-tubulin and other substrates.^8,9

The large number of available X-ray structures of HDAC6-bound inhibitors are consistent with the “three components” paradigm, where histone deacetylase ligands can be divided into distinct regions, namely, the (1) zinc binding group (ZBG), the (2) linker and the (3) cap (Figure 1).

Figure 1.

Representative examples of selective HDAC6 inhibitors including the two internally developed HDAC6 inhibitors, ITF3756 and ITF3985. For all inhibitors, it is possible to distinguish the ZBG in red, the spacer in blue, the cap in green, and the ramification in black.

The ZBG must satisfy several geometric restraints and trigger an interaction network with the residues of the catalytic core. The hydroxamic moiety fits this profile and represents the most explored ZBG motif. Although a large number of hydroxamate-based HDACis have been described, the diversity of the reported linker scaffolds is relatively low. Both aliphatic and arylic moieties are allowed in this position, but the latter are more successful because of their ability to fit the crevice between two conserved phenylalanines,^8,9 thus triggering a three-body π–π interaction. The conservative design of the ZBG and the linker areas has focused attention on the cap region, where the differences between HDAC isoforms are more significant. X-ray crystal structures have revealed some unique features of HDACis bound to HDAC6. (1) The hydroxamate group frequently exhibits monodentate binding with some relevant exceptions where the ZBG chelates Zn²⁺ in a bidentate fashion.^10,11 During monodentate coordination, the ionized hydroxyl group of the hydroxamate is directly bound to the zinc ion, and the carbonyl group interacts with the metal ion via a water molecule.^10,12 (2) Saturated, partially unsaturated, and aromatic cyclic linkers exhibit selectivity versus HDAC6, showing entropy-driven binding.¹³ The cap of nonbranched inhibitors points toward loop L1, whereas the binding conformer of the branched ligands interacts with both loops L1 and L2.

The most frequently observed binding mode of unbranched inhibitors has allowed the detection of a hydrophobic pocket involving the aliphatic side chains of Leu712 (loop L7), His463 and Pro464 (loop L1), and Phe583 (unless otherwise stated, the numbering of the residues refers to zHDAC6), the last of which is also usually engaged in a π–π interaction with the aromatic linker close to the metal ion. Most recent investigations have indicated this region as pocket L1, whereas loop L2 has been replaced by pocket L2, which involves residues Asn530 and Ser531 of loop L2 and Phe642 and Phe643 of loop L5 (Figure 2). Although one of the branches of inhibitors with nonlinear capping groups is frequently heavily influenced by monomers of the adjacent unit cells, smaller ligands such as ACY-1083 (Figure 1) orient the aromatic branch into pocket L1, whereas the aliphatic six-membered ring is positioned in pocket L2 (Figure 2). The larger compound NR-160 (Figure 1) bound to zebrafish CD2 (zCD2) exhibits an analogous binding geometry, where the short aromatic branch lies in pocket L1 and the longer chain of the 2,5-disubstituted tetrazole fits the L2 pocket (Figure 2).

Figure 2.

ACY-1083 (purple) and NR-160 (green) in the zHDAC6-CD2 binding site. The Zn ion interactions with zHDAC6 and the inhibitor ZBG are shown as thin gray lines. The picture shows the L1 pocket in terms of the Connolly¹⁴ surface. The light red surface indicates a relatively large crevice (pocket L1), involving Leu712, His463, Pro464, and Phe583; the light green surface indicates the L2 pocket, involving Asn530, Ser531, Phe642, and Phe643.

Relative to the isoforms of class I, HDAC6 selectivity has been explained in terms of loop L1 shape and position, forming a larger empty space close to the entrance of the active site cleft. Selective inhibitors can dock a bulky substituent efficiently onto only HDAC6, whereas not enough room is available in the other isoenzymes, leading to steric clashes.

A role in HDAC6 selectivity must be played by Ser531 (Ser568 in hHDAC6, Figure S3) in loop L2: all HDAC isoforms exhibit indeed an aspartic acid in that position except for HDAC6 (Asp to Ser), HDAC10 (Asp to Ala), and HDAC11 (Asp to Asn).

Herein, we investigate the role of fluorination in the selectivity of benzohydroxamate-based HDAC6 inhibitors and describe the X-ray crystal structures of fluorinated and nonfluorinated compounds (Figure 1) bound to the active site of zHDAC6 CD2 to identify the structural features that can be used in the design of novel, highly selective HDAC6 inhibitors.

The selective inhibition of hHDAC6 of compounds ITF3756 and ITF3985 (Figure 1) has been previously shown.¹⁵ Our enzymatic results (see Tables S1 and S2) confirmed what has been already reported by Hai and Christianson⁸ and Miyake et al.,⁹ i.e., the weak enzymatic activity of CD1 and the use of zCD2 as a valid surrogate of the actual drug target, the human HDAC6 CD2.

Alignment of the zCD2–ITF3756 (1-H) complex (1.90 Å resolution) with the ligand-free protein (Protein Data Bank (PDB) code: 5EEM(8)) shows an rmsd value of 0.49 Å (353 C_α involved), indicating no major conformational changes. Two monomers are present in the asymmetric unit, and both ligand geometries are equivalent (for more crystallographic details, see the Supporting Information).

The benzohydroxamate is stacked in the crevice made by Phe583 and Phe643, triggering a π–π interaction, whereas the ZBG exhibits a monodentate geometry (Figure 3A and Figure S2), which is in accordance with a large number of zCD2-inhibitor complexes.^7,10,12 The distance between the methylene moiety of the benzyl substructure and the Ser531 side chain (3.45 Å) is compatible with a C–H···O hydrogen bond (HB).¹⁶ The five-membered ring (1,5-disubstituted tetrazole) is mainly engaged in a hydrophobic π-alkyl¹⁷ interaction with Leu712 and a glycerol molecule, a component of the crystallization buffer. The apical thiophenyl moiety triggers a π–π T-shaped bond with Phe583 where the distance between the ring centroids is 5.44 Å, slightly longer than the 4.9–5.0 Å range determined from the gas phase calculation by using the Møller–Plesset perturbation theory and the Coupled Cluster method.¹⁸ There is also a π-alkyl interaction with the side chain of Pro464.

Figure 3.

ITF3756 (A) and ITF3985 (B) in zHDAC6 CD2. The Zn ion is represented as a gray ball. HBs (green dashed lines), π–π interactions (magenta dashed lines), and a π-alkyl interaction (purple dashed line) are recognized.

ITF3985 (2-FF) in zCD2 (2.30 Å resolution) aligned with the unbound enzyme exhibits a slightly larger rmsd (0.52 Å) compared to that of ITF3756. Only one monomer is present in the asymmetric unit (Table S3), and the benzyl hydroxamate moiety shares several common features with ITF3756. The hydroxamate group reveals monodentate coordination to Zn²⁺ with the sp³ oxygen directly bound to the metal ion (Figure 3B and Figure S2), whereas a water molecule directly binds the Zn²⁺ and makes HBs with Asp612, His573, His574, and the carbonyl oxygen atom of the hydroxamate (O···O distance equal to 2.81 Å). The phenyl substructure is sandwiched between Phe583 and Phe643 with the usual π–π stacking interaction. The fluorine atom that faces the loop L2 residues apparently interacts (as determined by Discovery Studio Visualizer 2020, automatic noncovalent bond recognition) with the C_β of Ser531 and the C_α of Gly582 via a C–H···F weak HB. It should be of note that HBs involving fluorine atoms are controversial from both the experimental and theoretical points of view.^19,20 Despite the poor propensity of the fluorine atom to engage in noncovalent bonds (halogen bonds are rarely reported),²¹ there is evidence of a possible weak noncovalent interaction between the small electronegative halogen and the backbone’s carbonyl moiety.²² Indeed, several X-ray structures have reported F···C=O distances²³ in the 3.0 Å −3.7 Å range, even if the details of such bonds are still unclear.²⁴

The measured distances suggest a possible interaction between the fluorine atom and the C_α of Gly582 (3.77 Å), whereas the C_α of Ser531 is more distant (4.67 Å). A similar situation was observed in the Bavarostat zCD2 complex (PDB code: 6DVO), where the distances from Gly582 and Ser531 are 3.63 and 5.09 Å, respectively (Figure 4).

Figure 4.

Bavarostat (orange carbon atoms) superimposed with the crystal structure of compound 2-FF (magenta carbon atoms) in zCD2. The fluorinated ring of Bavarostat penetrates deeper into the catalytic tunnel, showing bidentate hydroxamate-Zn²⁺ coordination.

The bridging methylene motif of ITF3985 and the Ser531 oxygen atom are 4.22 Å apart, showing no evidence of hydrogen bonding. This loss of a HB, which is present in the HDAC6-ITF3756 complex, is probably compensated by the weak C–H···F HB and an attractive interaction between the serine side chain and the tetrazole of ITF3985; the position of the heterocycle shows two nitrogen atoms with lone pairs in the plane of the ring facing loop L2. Although the O···N distance (3.03 Å) is larger than the hydrogen bond donor–hydrogen bond acceptor (HBD-HBA) cutoff in Discovery Studio, the maximum distance between heavy atoms in weak HBs (whose electrostatic component is higher) is 3.2–4.0 Å.²⁵ Comparison of the positions of the serine residues in the ITF3756 and ITF3985 complexes with the enzyme indicates a displacement (0.30 Å) of the residue center of mass (Figure 5A) that pushes the tetrazole moiety of ITF3985 closer to the serine side chain, suggesting a possible contribution of the heterocycle to the binding energy.

Figure 5.

(A) Overlay of ITF3756 (1-H, green carbon atoms) and ITF3985 (2-FF, magenta carbon atoms) complexed with zCD2. The centroids of the phenyl substructures are not superimposable; the fluorinated aryl moiety lies 0.45 Å closer to the Zn ion. (B) Overlay of ITF3756 (green carbon atoms) and Bavarostat (orange carbon atoms) (PDB code 6DVO) complexed with zCD2. The center of mass displacement is larger (0.76 Å) with the monofluorinated inhibitor buried closer to the metal ion.

The second fluorine atom on the benzyl hydroxamate substructure is a unique feature among zCD2-inhibitor complexes available in the literature. Distance analysis reveals a possible contact (3.54 Å) between the halogen atom and the imidazole nitrogen atom of the His614 side chain. Although halogen bonds between fluorine and heteroatoms are rare,²¹ the measured distance allows speculation of a possible weak interaction.

The tetrazole moiety of the inhibitor exhibits a π–π bond with Phe583 (T-shaped geometry), whereas the pyrimidine moiety shows similar behavior from the thiophenyl substructure in ITF3756, suggesting a π–π bond with the His463 imidazole and a π-alkyl interaction with the Pro464 side chain (L1 pocket). The long flat cap structure allows the inhibitor to reach more distant portions of L1 and come close to Asp460, a key residue in most recent HDAC6 hydroxamate-based inhibitors, confirming the docking prediction¹⁵ and most recent structural results.²⁶

Fluorination of benzohydroxamate-based HDAC6 inhibitors has been observed to enhance either their potency^27,28 or their selectivity over HDAC1.^10,15 To clarify the role of fluorination, we prepared and tested the fluorinated analog of ITF3756 (1-FF) and the nonfluorinated analog of ITF3985 (2-H) (see Table 1) against HDAC1 and HDAC6. Unfortunately, good-resolution crystal structures of zCD2 with compounds 1-FF and 2-H have not yet been obtained.

Table 1. Role of Fluorination: Comparison of Nonfluorinated Compounds 1-H (ITF3756) and 2-H with their Corresponding Fluorinated Analogs 1-FF and 2-FF (ITF3985).

^aEnzymatic data (IC₅₀) in nM are the mean of at least two experiments obtained from the curve fitting of a five-point enzymatic assay starting from 100, 30, or 10 μM with 10-fold serial dilutions (technical triplicate).

^bExperiments were performed in technical triplicates (single experiment). SDs were calculated on technical replicates.

^cSelectivity indicator, expressed in terms of the ratio of IC₅₀ values (HDAC1/HDAC6). Assays were performed using fluorogenic Fluor de Lys as the substrate. HDAC1 and HDAC6 were used at a catalytic concentration of 2 nM and 500 pM respectively.

In contrast with previous reports,^27,28 our results demonstrate that fluorine atoms do not significantly improve the activity against isoform 6 but rather reduce the potency on HDAC1 with the effect of enhancing the selectivity. Fluorinated analogs also maintain good selectivity toward the other isoforms, even though some loss over class II HDACs must be acknowledged (see Table 2).

Table 2. Complete Isoform Inhibition Profile of Compounds 1-H, 1-FF, 2-H, and 2-FF.

Compd	1-H (ITF3756)	1-FF	2-H	2-FF (ITF3985)
HDAC1 IC50a	924 ± 128	7093b ± 280	930 ± 166	3960 ± 689
HDAC2 IC50a	3698 ± 487	9232b ± 4073	3007b ± 342	17196 ± 761
HDAC3 IC50a	865 ± 134	6045 ± 170	884 ± 176	5179 ± 843
HDAC4 IC50a	3068 ± 291	660b ± 65	4601b ± 212	562b ± 19
HDAC5 IC50a	2035 ± 722	699b ± 88	4999b ± 192	2788b ± 12
HDAC6 IC50a	17 ± 8	19 ± 6	7 ± 2	5 ± 1
HDAC7 IC50a	985 ± 407	472b ± 75	1010 ± 194	727 ± 161
HDAC8 IC50a	984 ± 368	996b ± 101	320b ± 26	504 ± 157
HDAC9 IC50a	1439 ± 363	658b ± 83	1705b ± 62	480b ± 20
HDAC10 IC50a	1695 ± 391	8740b ± 2765	474b ± 97	9078b ± 413
HDAC11 IC50a	1015 ± 378	5662b ± 3719	762b ± 373	3071b ± 1120

^aEnzymatic data (IC₅₀) in nM units are the mean of at least two experiments obtained from curve fitting of a five-point enzymatic assay starting from 100, 30, or 10 μM with 10-fold serial dilution (technical triplicate).

^bExperiments were done in technical triplicate (single experiment). SDs were calculated on technical replicates. Assays were performed using fluorogenic Fluor de Lys and N_ε-trifluoroacetyl-l-lysine as substrates. HDACs from 1 to 11 were used at a catalytic concentration of 2 nM, 3 nM, 400 pM, 50 pM, 700 pM, 500 pM, 50 pM, 5 nM, 1 nM, 35 nM, and 25 nM.

Recent findings have raised doubts about the reliability of classical inhibition assays as regards HDAC10, HDAC11, and class IIa HDAC isoforms (i.e., HDAC4, HDAC5, HDAC7, and HDAC9) due to the uncertainty on their correct substrates.¹ Accordingly, our discussion will mainly focus on the selectivity of our HDAC6 inhibitors over class I HDAC isoforms with high deacetylase activity (i.e., HDAC1, HDAC2, and HDAC3).

To gain a deeper understanding on the role of fluorination in the selectivity of the benzohydroxamate-based structures, the halogen contribution has also been assessed on a sequence of short substituted benzohydroxamates along with their close analogs that consider the bioisosterism between pyridine/pyrimidine and mono/difluorinated benzene (for more details about their synthesis, see the Supporting Information).²⁹ The results shown in Table 3 confirm the intrinsic selectivity versus HDAC6 of small hydroxamic inhibitors having a cyclic linker¹³ and reveal interesting trends in the potency and the selectivity driven by the presence of N and F in specific positions. All these small molecules indeed show an enhancement in both potency and selectivity proportional to the number of substitutions (achieved by either fluorination^15,27 or C-replacement with nitrogen).³⁰ This trend is true for 2,5-disubstituted pyridines and pyrimidines and 3-mono-²⁷ and 3,5-difluorinated¹⁵ phenyl scaffolds (compare 3-H with 3-F and 3-N; 4-H with 4-N, 4-F, and 4-FF; 5-H with 5-N, 5-NN, and 5-FF; and 6-H with 6-F; Table 3) and apparently correlates with the decrease in the pK_a values of the hydroxamic acid (see Table S4).^27,28

Table 3. Role of Fluorination: Inhibitory Potencies and Selectivities of Short Substituted Benzohydroxamates and Close Analogs^c.

^aEnzymatic data (IC₅₀) in nM are the mean of at least three experiments obtained from the curve fitting of a five-point enzymatic assay starting from 100, 30, or 10 μM with 10-fold serial dilutions (technical triplicate).

^bSelectivity indicator, expressed in terms of HDAC1/HDAC6 IC₅₀ ratio. Assays were performed using fluorogenic Fluor de Lys green as the substrate. HDAC1 and HDAC6 were used at a catalytic concentration of 1 nM and 300 pM, respectively.

^cGivinostat (7H) has been Added as a Representative Benzohydroxamate-Based pan-HDACi.

Finally, the same strategy has been applied to the broad-spectrum benzohydroxamate-based HDACi givinostat (for more details about the synthesis, see the Supporting Information), and the results confirmed a 10-fold enhancement in selectivity versus HDAC1 achieved by fluorination along with a nonsignificant increase in the potency (compare 7-H with 7-FF, Table 3). As a general rule, the cap moiety and its interactions with rim residues seem to undo the gain in potency achieved by the fluorination.

The three-dimensional structures of zHDAC6-CD2 in complex with ITF3756 (1-H) and ITF3985 (2-FF) are consistent with the observed inhibition behavior; the overlay of the inhibitor-enzyme structures of the nonfluorinated compound and the fluorinated one (Figure 5A) shows, indeed, a similar position of the benzyl substructure in the catalytic channel, although a small shift is detected between their phenyl centers of mass (0.45 Å, Figure 5A) with the centroid of ITF3985 closer to the zinc ion. A similar picture (Figure 5B) is also observed for the Bavarostat zCD2 complex (6DVO), where the drift of the aromatic ring is even more pronounced (relative to ITF3756, the centroid is 0.75 Å closer to the zinc ion). Several factors are likely to contribute to the shift of the benzohydroxamate moiety, and the most relevant in order of importance are: the interactions between the inhibitor and the enzyme, the crystal packing of the unit cells, and the pK_a of the hydroxamic acid.

As anticipated earlier in the text, the serine at the entrance of the catalytic channel (Ser531) of HDAC6 plays a role in the selectivity. Sequence alignment of all hHDACs (Figure S3) clearly indicates a serine residue as a unique feature of isoform 6, whereas almost all other family members show an aspartate on loop L2.

HDAC6’s serine occupies a smaller volume at the catalytic tunnel entrance and exhibits a lower electron density. This locally reduced electron density along with the larger tunnel entrance and possible C–H···O HB between the serine side chain and the methylene group of this class of inhibitors seem to be the reasons for their remarkable selectivity over class I HDAC isoforms. A fluorine atom oriented toward loop L2 is probably penalized by Coulombic and steric repulsion in all HDACs with an aspartate at the tunnel entrance.

Superposition (see Figure S4) of the HDAC6–2-FF complex on HDAC1 and HDAC7 (PDB codes 5ICN and 5ZNR, respectively) indicates that the inhibitor binding mode detected in HDAC6 (L1 pocket) is forbidden in class I and class IIa isoforms because of the larger volume of the key residue aspartate, which partially obstructs the entrance of the catalytic tunnel. On the other hand, docking simulations reveal the presence of an alternative narrow and deep hydrophobic pocket between loops L7 and L5 just for HDAC7 and not for HDAC1 (see Figure S5) that can fit the cap of the fluorinated compounds and explain the enhancement in their potency. The equivalent region in HDAC1 is too shallow and does not allow a suitable fit of the cap (see Figure S5). In this docking pose, the fluorine atom of 1-FF and 2-FF lies at a higher distance from the aspartate, but the phenyl moiety still interacts with the aromatic crevice of the catalytic tunnel.

In summary, our results show that fluorine atoms can increase binding energy due to the π–π interaction in the aromatic crevice of the catalytic tunnel of HDAC6¹³ and the additional weak interactions between the halogen atom and the key residue Ser531. However, these effects translate into better potency on HDAC6 only for small cap-less inhibitors and not for larger molecules like 1-FF and 2-FF, where a more complex pattern of interactions is present. On the other hand, the halogen atoms penalize the ligands fit in class I isoforms and partially restore the affinity for class IIa isoforms, leading to an enhanced selectivity over class I without significant loss of selectivity over class IIa. Taken together, these findings contribute to rationalizing important structural features that can be used in the design and development of novel HDAC6 inhibitors with a wider therapeutic window.

Glossary

Abbreviations

NCI

noncovalent interaction

rmsd

root-mean-square deviation

CD1

catalytic domain 1

CD2

catalytic domain 2

HDAC

histone deacetylase

z-HDAC6

zebrafish histone deacetylase isoform 6

ZBG

zinc binding group

HB

hydrogen bond

HBD

hydrogen bond donor

HBA

hydrogen bond acceptor

CCP4

collaborative computational project number 4

DMSO

dimethyl sulfoxide

PEG

polyethylene glycol

SOLEIL

Source optimizée de lumière d’énergie intermédiaire du LURE (Laboratoire pour l’utilization du rayonnement électromagnétique)

HEPES

4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid

8xHis

PolyHis Tag

MBP

Maltose-Binding Protein

TEV

catalytic domain of nuclear-inclusion-a endopeptidase from tobacco etch virus

Ni-NTA

Ni²⁺-nitrilotriacetic acid agarose

TCEP

tris(2-carboxyethyl)phosphine

PDB

Protein Data Bank

WT

wild type

prep HPLC

preparative HPLC

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00425.

• Synthesis of compounds; description of enzymatic tests on recombinant hHDAC6 mutants Y386F and Y782F, measurements of their specific activities and K_m and comparisons with hHDAC6 ones (Table S1), measurements of IC₅₀ values of some HDACis (Table S2); method for enzyme inhibitory activity assay, for protein production and purification, for crystallization and data collection, and for structure determination; diffraction data and refinement statistics (Table S3); comparison of the crystal packing contacts between zHDAC6-ITF3985 and zHDAC6-Bavarostat complexes (Figure S1); details of the hydroxamate denticity of 1-H and 2-FF (Figure S2); alignment of 11 hHDACs (Figure S3); comparison between HDAC6 binding mode and class I and IIa (Figure S4); docking results for ITF3985 in HDAC1 and HDAC7 (Figure S5); calculated pK_a values (Table S4) (PDF)

Accession Codes

The atomic coordinates and crystallographic structure factors of HDAC6 complexes with inhibitors 1-H and 2-FF have been deposited in the Protein Data Bank (www.rcsb.org) with accession codes 7O2P and 7O2R, respectively. The authors will release the atomic coordinates and experimental data upon article publication.

Author Contributions

^§ G.S. and C.D.C. contributed equally to this work. C.S. and A.S. supervised the project. G.S. and C.D.C. assembled the manuscript. B.V. and M.M. synthesized, purified, and characterized the compounds discussed in this paper. K.Z. produced zHDAC6-CD2 and crystallized the enzyme–inhibitor complexes. C.D.C. and K.Z. solved, refined, and analyzed the X-ray structures with the help of G.S. G.C. assayed the compounds. G.F. reviewed the manuscript.

The authors received funding from Regione Lombardia, Project E79J19000480007 - ID 1165235, as part of the “European Regional Development Fund (ERDF) of the Regional Operational Program (ROP) 2014–2020”. Cyprian D. Cukier and Karol Zrubek received financial support from Italfarmaco for the crystallization of ITF3756 and ITF3985.

The authors declare the following competing financial interest(s): Cyprian D. Cukier and Karol Zrubek received financial support from Italfarmaco for the crystallization of ITF3756 and ITF3985.