Appendix

Table S1. Epigenetic mechanisms and their respective involved protein families and domains Epigenetic

mechanism Writers Erasers Reader domains Refs

DNA

SUMOylation SUMO ligases SUMO-specific proteases

SUMO-interacting motif

176

Abbreviations: plant homeodomain (PHD); malignant brain tumor (MBT); zinc-finger conserved cysteine and tryptophan (ZF-CW); protein domains that contain repeated tryptophan and aspartic acid motifs over 40-60 residues (WD40); proline-tryptophan-tryptophan-proline domain (PWWP); BRCA1 (breast cancer susceptibility gene 1) C-terminal (BRCT); small ubiquitin-like modifier (SUMO).

38

Table S2. HDAC6 crystallographic structures available on the Protein Data Bank (PDB)^a.

PDB code Species HDAC6 corresponding domain Date of release Refs

5EDU Human Catalytic domain 2 (CD2) 2016 ¹²¹

5GOG Zebrafish Catalytic domain 1 (CD1) 2016 ¹²⁴

5EEF Zebrafish Catalytic domain 1 (CD1) 2016 ¹²¹

5GOH Zebrafish Catalytic domain 2 (CD2) 2016 ¹²⁴

5EEI Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EEK Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EEM Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EEN Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EF7 Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EF8 Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EFB Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EFG Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EFH Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EFJ Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EFN Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5EFK Zebrafish Catalytic domain 2 (CD2) 2016 ¹²¹

5GOI Zebrafish Catalytic domain 1 and catalytic domain 2

(CD1 and CD2) 2016 ¹²⁴

5GOJ Zebrafish Catalytic domain 1 and catalytic domain 2

(CD1 and CD2) 2016 ¹²⁴

3C5K Human Ubiquitin-binding domain (ZnF-UBP) 2008 ¹⁷⁷

3PHD Human Ubiquitin-binding domain (ZnF-UBP) 2011 ¹¹⁰

3GV4 Human Ubiquitin-binding domain (ZnF-UBP) 2009 ¹⁷⁷

5KH3 Human Ubiquitin-binding domain (ZnF-UBP) 2016 ¹⁷⁸

5B8D Human Ubiquitin-binding domain (ZnF-UBP) 2016 ¹⁷⁸

5KH7 Human Ubiquitin-binding domain (ZnF-UBP) 2016 ¹⁷⁸

5KH9 Human Ubiquitin-binding domain (ZnF-UBP) 2016 ¹⁷⁸

5WBN Human Ubiquitin-binding domain (ZnF-UBP) 2017 ¹⁷⁸

5WPB Human Ubiquitin-binding domain (ZnF-UBP) 2017 ¹⁷⁸

5GOF Zebrafish Ubiquitin-binding domain (ZnF-UBP) 2016 ¹²⁴

a: accessed on Nov. 8^th 2017.

39

Table S3. HDAC6 partners and their demonstrated interacting domain(s)

Partners Related function Demonstrating methods Interacting

domains Refs

Tat Tat deacetylation (Lys28):

HIV transactivation

Pull down assay with WT and mutated ZnF-UBP

Chapter 2.

HDAC inhibitors from diverse chemical

libraries: looking for HDAC6 selectivity

41

2.1 Introduction

The development of a new drug is a long multi-step process that can lasts from ten to twenty years. It usually initiates because of an unmet clinical need as an underlying driving motivation to justify a drug discovery effort (Fig. 9). The very first phase of such an adventure is the target identification (proteins, genes, RNA) and validation, based on basic research, for which disease correlation and

“druggability” have been assessed and demonstrated (validated target). Hits generation phase immediately follows. Hits are compounds identified during screening campaigns (enzymatic, cellular or tissue assays), which have shown the desired activity and whose activity profile has been established through repeatable experiments ^1-3. There are many ways for towards hits generation, from high throughput screenings (HTS, i.e. systematic testing of entire large libraries) to more knowledge-based techniques like focused library screening or computer-aided virtual screening (VS).

Figure 9. The drug discovery process. All key phases are given with the corresponding approximate duration time. The hit generation step is highlighted, showing the different strategies applied. This figure is adapted from Dickson and Gagnon, Bleicher et al., and Hughes et al. works and constructed using the images bank provided by Servier Medical Art and under a Creative Common Attribution 3.0 unported license ^1-3. ID: identification, NDA: new drug application.

The term “virtual screening” has been used since the 1990s and refers to hit compounds identification through computational algorithms and models ^4-7. Molecular docking was the first virtual screening method described in 1982 by Kuntz and co-workers as a geometric approach to investigate interactions between a macromolecule and a ligand ⁸. Molecular docking comprises two linked steps which are: i) a search algorithm that explores the energy landscape of a ligand into a pocket through a conformational search; and ii) a scoring function that evaluates the interactions resulting from such a conformational search ^5,9-12. At first, this computational method was seen as a time- and cost-saving method in comparison with the classical HTS. Nowadays, it is used in combination with HTS to increase

42

the success rate of the screening campaigns through activity-enriched libraries and structure novelty

1,2,4,5,13-15.

In this chapter, virtual screening methods combined with classical enzymatic in vitro screening will be used in order to identify new HDAC6 inhibitors. Usually, HDAC inhibitors (HDACi) match the following pharmacophore: a cap group, to interact at the enzyme’s channel rim, a zinc-binding group linked to the latter part by a hydrophobic chain that fits into the channel. Five chemical classes were screened for HDAC6 inhibition, known for being able to interact with metalloproteins through metal coordination: chalcones, aurones, 1,3,4-oxadiazoles, pyrimidinetriones, and thiosemicarbazides.

Chalcones, also called 1,3-diphenylpropenones or benzylidene-acetophenones, are a subclass of naturally occurring compounds, and belong to the flavonoid family. They are easily substituted with various groups, from halogens to hydroxyl and alkyloxy moieties ¹⁶. Hydroxylated chalcones have been tested on SIRT1, SIRT3 and class I HDACs, revealing IC50 under 100 µM ^17-19. To go further with natural products as a source of potential HDAC6 inhibitors, aurones could be interesting compounds.

They are isomers of flavones and derivatives of chalcones. Their scaffold consists of a benzofuranone ring linked through a carbon-carbon double bond to a phenyl ring. Aurones have a broad therapeutic potential, displaying, in particular, affinities for Aβ aggregates as imaging agents and inhibitory activities against acetylcholinesterase in Alzheimer’s disease diagnosis and treatment ^20,21. Moreover a 5’-bromo-6’-hydroxyaurone displayed SIRT1 full inhibition at 10 µM ²².

Within the chemical diversity of compounds prone to interacting with HDACs, 1,3,4-oxadiazoles are heterocyclic compounds containing a five-membered ring with one oxygen and two nitrogens. There are three isomers (1,2,4-; 1,2,3- and 1,2,5-) but 1,3,4-oxadiazoles are one of the most studied because of their many biological activities (from antibacterial to anti-diabetics), and because they are bioisosters of carboxylic acids, esters, and carboxamides ²³. They have been used for targeting HDAC1, and as isosters of alkylamides against HDAC4 even if their binding mode was not described

24,25. Pyrimidinetriones are molecules containing a six-membered ring with two nitrogen alternating with three ketone moieties. They have been shown to chelate the zinc ion of MMPs (matrix metalloproteases) and of the TNFα (tumor necrosis factor α) converting enzyme, in an effort to find non-hydroxamate inhibitors ^26-28. Finally, thiosemicarbazides were successfully reported as zinc chelators, making complexes through a distorted trigonal bipyramid geometry, and, in particular, as metallo-β-lactamases inhibitors (zinc-containing metalloenzymes responsible for β-lactam antibiotics resistance) ^29,30.

The strategy was the following: a series of aurones were first tested in vitro against HDAC class I isozymes and HDAC6, revealing biological activities rationalized through molecular modeling.

Results are described in the section 2.2 and experimental details are given in the section 2.8. To further explore the pharmacochemistry of the aurone scaffold, a VS campaign was also designed, using both HDAC2 and HDAC6 structures as receptors. A series of 1,3,4-oxadiazoles was added to the ligands set, so the ligand set reached one hundred molecules. Best-ranked compounds were then selected for IC50

determination on HDAC6 and HeLa nuclear extract as a source of HDAC class I enzymes. Results are given in the section 2.3. In parallel, a small set of chalcones, pyrimidinetriones, and thiosemicarbazides, were directly tested in vitro for HDAC6 inhibition and results rationalized in silico. Results are given in the section 2.3 and experimental procedures in the section 2.8.

43