Selective pharmacological inhibitors of HDAC6 reveal biochemical activity but functional tolerance in cancer models
Introduction
Histone deacetylases (HDACs) are a class of enzymes respon- sible for altering the acetylation status of target proteins by removing acetyl groups from side-chain acetylated lysine (K)- residues.1 There are four classes of HDACs, each containing different isozymes, divided according to their homology to their yeast counterparts, cellular localization and substrate specificity: the zinc-dependent class I (HDAC1-3 and 8), class II (HDAC4-7 and 9–10) and class IV (HDAC11) and the nic- otinamide adenine dinucleotide (NAD+)-dependent class III or sirtuins.2,3 Class I HDACs reside predominantly in the nucleus where their main substrate, histones, can be found. Class II enzymes can shuttle between the nucleus and cyto- plasm and HDAC6 is located almost exclusively in the cyto- plasm. HDAC11 bears the closest resemblance to HDAC3 and 8 and resides mainly within the nucleus.2,3 Sirtuins differ from the classical HDAC family, requiring NAD+ for their activity instead of Zn2+.
Key words: histone deacetylase 6, inhibition, Tubathian A, Tubastatin A, tumour
Since HDACs are often overexpressed in cancer and are believed to play an essential role in carcinogenesis and cancer progression, they are promoted as promising therapeutic tar- gets. Capitalizing on this, numerous HDAC inhibitors (HDACi) have been designed as clinical agents, leading to the FDA approval of four HDACi for the treatment of specific types of T-cell lymphoma and multiple myeloma: Vorinostat (SAHA), Romidepsin, Panobinostat and Belinostat. HDACi act through complexation of the essential Zn-kation, present in the catalytic site of the enzymes, thus inhibiting the deace- tylase activity. They assert their biological effect by inducing apoptosis or cell-cycle arrest, reducing angiogenesis and/or metastasis and enhancing anti-tumour immunity.4–6 On the downside, the majority of HDACi are non-selective, implying that they inhibit multiple HDAC isozymes, and display toxic side effects due to their broad activity.7 In light of this, recent research focusses on the development and implementation of isoform-selective HDACi.
HDAC6 is mainly localized in the cytoplasm, due to its nuclear export signal and cytoplasmic retention domain.8 Con- sistently, it performs its function on a range of cytoplasmatic proteins, such as α-tubulin, cortactin and HSP90.9,10 Due to its interaction with proteins involved in cell growth, apoptosis, migration, protein degradation, HDAC6 is considered an important therapeutical target for cancer treatment.10–12 Fur- thermore, since mice lacking HDAC6 develop normally, minor to no side-effects are believed to be expected as a result from HDAC6 inhibition.13 The first selective HDAC6i, Tubacin, showed excellent activity and good selectivity, but displayed poor druglikeness due to its high lipophilicity.14 Numerous HDAC6i have been developed since, with Tubastatin A being one of the most active compounds available to date.15 Recently, our group designed a series of sulphur-analogues of the latter compound, called Tubathians, which show excellent enzymatic and cellular selectivity and activity profiles against HDAC6 and possess good ADME-Tox properties.16,17
The selective inhibition of HDAC6 has been claimed to result in a variety of anti-cancer properties such as a reduction in cell growth, motility and angiogenesis.11 Two HDAC6i, Ricolinostat and ACY-241, are currently undergoing clinical trials against several cancer types, either alone or in combination with existing drugs. However, the interpretation of these results are hampered because of the simultaneous inhibition of other HDAC isozymes, due to the low selectivity of certain compounds or the high con- centrations used in in vitro and in vivo assays (Supporting Infor- mation Tables 1 and 2). This implies the need for highly selective HDAC6i that should be used at selective concentrations to elucidate the exact function of HDAC6’s deacetylation activity in cancer progression. In our study, we aimed to address this challenge by assessing the impact of Tubathian A, in comparison to other HDAC6i Tubastatin A, Tubacin and Ricolinostat and the non-selective HDACi Vorinostat on metastasis-associated cellular activities in vitro and in vivo. In light of several literature claims on the potential application of selective HDAC6i in the field of oncology, an unambiguous and clear-cut assessment of the interaction between HDAC6 inhibition and anti-cancer effects will certainly contribute to the advancement of HDACi- based cancer therapy.
Materials and Methods
Compounds
HDAC6 inhibitor Tubathian A (8A) was synthesized as described previously.16 The purity of this compound exceeded 99% as measured by 1H-NMR and LC–MS analysis. HDAC inhibitors Tubastatin A (Tub A), Tubacin (Tub), Ricolinostat (ACY) and Vorinostat (SAHA) were purchased from Selleckchem (Munich, Germany). All compounds were dissolved in dimethyl sulfoxide (DMSO) and stored at −20◦C.
Cell culture
SK-OV-3 is a human ovarian cancer cell line obtained from ATCC (HTB-77), which was made luciferase positive (SK-OV-3-Luc) as previously described.18 4 T1-Luc mouse breast cancer cells were purchased from Sibtech (Brookfield,USA) and selected with Zeocine (500 μg/mL, Invitrogen).
MCF-7 and MDA-MB-231 human breast cancer cells and human mesothelioma MSTO-221 h cells were purchased from ATCC (HTB-22, HTB-36 and CRL-2081 respectively). Human colorectal adenocarcinoma cell line HT-29-Luc were kindly provided by Dr. Hackl (University of Toronto).19 HAP1 cells are near-haploid fibroblast-like human cells derived from the chronic myelogenous leukemia cell line KBM-7. The HDAC6 knockout variant HAP1-KO was edited by CRISP/Cas9 to contain a 1 bp insertion in the coding exon for HDAC6. Both were purchased form Horizon Discovery (Cambridge, UK). All cell lines, except HAP1 and HAP1-KO, were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM),supplemented with 10% foetal bovine serum, 100 U/mL penicillin, 100 μg/mL streptomycin and 2 μg/mL fungizone (Life Technologies, Ghent, Belgium). HAP1 and HAP1-KO cells were cultured in Iscove’s Modified Dulbecco’s Medium (IMDM) supplemented with 10% foetal bovine serum,100 U/mL penicillin, 100 μg/mL streptomycin and 2 μg/mL fungizone (Life Technologies). Cells were maintained at 37◦C and 10% (SK-OV-3-Luc) or 5% (all others) CO2 in air. All cultures were monthly tested for Mycoplasma by using MycoAlert Plus Kit (Lonza, Basel, Switzerland).
IncuCyte ZOOM monitored assays
Real-time monitoring of cell confluency and scratch wound migration was done using IncuCyte ZOOM System (Essen Bioscience, Hertfordshire, UK) according to the manufac- turer’s guidelines. For cell confluency monitoring, cells were seeded in Corning 96-well plates (1,000 cells/well, 100 μL/well) and allowed to adhere overnight at 37◦C and 5% CO2. Subse- quently, cells were treated as indicated and microscopic images (4 images/well) were taken every 2 h for the indicated time. All images were analysed and cell confluency was deduced
using IncuCyte software. For scratch wound migration, cells were seeded in 96-well plates (100 μL/well, IncuCyte Image- Lock Plates, Essen Bioscience) and allowed to reach 100% con- fluency at 37◦C and 5% CO2. A scratch wound was made using a WoundMaker (Essen Bioscience) according to the manufac- turer’s instructions and the cells were treated (100 μL) as indi- cated. The relative wound density was followed over time by taking microscopic pictures (1 image /well) every 2 h during 24 h, which were analysed using IncuCyte software. Each con- dition was performed at least in sixfold and each experiment was performed in triplicate.
xCELLigence real-time cell analysis of trans-well migration Real-time monitoring of cell trans-well migration was done using xCELLigence DP Real-Time Cell Analyzer (RTCA, Leusden, The Netherlands) according to the manufacturer’s guidelines and as described before.20 In the lower chamber of 16-well plates 160 μL medium with 10% FBS was added and subsequently 50 μL of serum-free medium was added in the upper wells. After 1 h of equilibration at 37◦C and 5% CO2, the background was measured. Next, SK-OV-3-Luc cells were seeded in the upper wells (100,000 cells/well, 50 μL) and treat- ment was added (as indicated, 50 μL). Each condition was performed in duplicate and the electrical impedance was measured every 5 min for 24 h by RTCA software. The trans-well migration was expressed as the cell index, which is a measure for the change in impedance at each time point. Each experi- ment was performed in triplicate.
In vivo SK-OV-3-Luc peritoneal metastasis model
Animal experiments were conducted in accordance with the local ethics committee (ECD 17/51, Ghent University Hospi- tal). Mice were housed in controlled conditions with constant temperature (26–28◦C) and humidity (40–60%). Food and water were provided ad libitum. Ten-week-old female Swiss nude mice (Charles River, l’Arbresle Cedex, France) were intraperitoneally injected with 1 × 106 SK-OV-3-Luc human ovarian cancer cells suspended in 250 μL of serum-free medium. The following day all mice were imaged for bioluminescence. Mice were given an intraperitoneal injection of 100 μL D-luciferin (Caliper Life Sciences, Hopkinton, USA) in PBS (150 mg/kg mouse) and were anesthetized with isoflurane (5% in oxygen for induction and 1.5% in oxygen for mainte- nance, IsoFlo, Abbott, Belgium). Imaging was initiated 10 min after injection using an IVIS Lumina II (Caliper Life Sciences). Exposure times were set automatically. Mice were randomized in three groups (N = 7 or 9) based on bioluminescent signal and treatment was started. Mice were treated daily with con- trol (5% DMSO in PBS further abbreviated as vehicle) or treatment (5 or 20 mg/kg 8A) conditions by intraperitoneal injection. Tumour development was assessed by weekly biolu- minescence imaging until 5 weeks after inoculation. The humane endpoint was set at the time that the mice showed clear signs of advanced carcinomatosis. Tumour tissue was extracted 1 h after final treatment and partly homogenized for protein extraction using a gentleMACS Dissociator (Miltenyi Biotec, Auburn, USA), according to the manufacturer’s guide- lines, and used for immunoblotting. Part of the tumour tissue was fixed in 4% formalin and embedded in paraffin. Sections of 4 μm were deparaffinised, hydrated and stained with hae- matoxylin and eosin (H&E). Sections of 2 μm were stained for Ki67 as previously described.21
In vivo 4 T1-Luc breast cancer model
Four-week-old female BALB/c mice (Charles River) were orthotopically injected in the mammary fat pad with 1 × 106 4 T1-Luc mouse breast cancer cells suspended in a 100 μL mixture of serum-free medium and Matrigel (1:1). Imaging for bioluminescence was performed the day after, mice were randomised in three groups based on bioluminescent signal (N = 6 or 7) and treatment was started. Mice were treated daily by intraperitoneally injection of control vehicle or treat- ment (5 or 20 mg/kg 8A) conditions. Five weeks post inocula- tion mice were given an intraperitoneal injection of 150 mg/kg D-luciferin and sacrificed 20 min after injection. Mammary tumour and lungs were resected for ex vivo bioluminescence imaging 1 h after final treatment. Part of the tumour tissue was homogenized for protein extraction using the gentleMACS Dis- sociator and used for immunoblotting. The remaining tumour tissue and lungs were fixed in 4% formalin and embedded in
paraffin. Sections of 4 μm were deparaffinised, hydrated and stained with haematoxylin and eosin (H&E).
Statistical analysis
Statistical analysis was performed using Graphpad Prism 5. Area under Curve (AUC), extracellular glutamine concentration, invasion index and metabolic activity were analysed using One- Way Anova with Dunnett’s Multiple Comparison Post Test. Mann–Whitney U test was used to compare bioluminescence and quantification of immunoblotting in in vivo assays. Pearson Chi-Square test was performed to compare the number of mice with lung metastasis. All tests were two-sided and p value <0.05 was considered statistically relevant. In figures, * represents p ≤ 0.05, ** p ≤ 0.01 and *** p ≤ 0.001. Additional methods Additional material and methods can be found in Supporting Information. Results Tubathian A is a highly active and selective HDAC6i Tubathian A was developed as a sulphur-analogue of Tubastatin A, a class IIb specific HDACi, and showed excellent activity against HDAC6 (IC50 = 1.9 nM) and only limited inhibition of other HDACs, especially those of class I (e.g. IC50 HDAC1 = 11 μM).16,22 Compared to literature on enzymatic HDAC inhibition by other HDAC6i (Tubastatin A, Tubacin and Ricolinostat) or non-selective inhibitors (Vorinostat), Tubathian A showed superior HDAC6 selectivity (Fig. 1a).15,23,24 When treating SK-OV-3-Luc ovarian cancer cells for 1 h with Tubathian A, a concentration-dependent rise in the K-40 acetylation level of α-tubulin, an important HDAC6 substrate, was observed starting from concentrations as low as 0.1 μM (Fig. 1b). Interestingly, acetylated histone 3, a nuclear HDACs substrate, showed increased K-9 acetylation when using ≥10 μM of Tubathian A, indicating that this compound, although being a highly selective HDAC6i, affects other HDACs at such high concentrations (Fig. 1b and Supporting Information Fig. 1A). Although slightly less pronounced, similar results were obtained for Tubastatin A and Tubacin. Ricolinostat resulted in a clear rise in acetylated histone 3 at a concentration as low as 1 μM, demonstrating its inferior selectivity profile. Treatment with the non-selective inhibitor Vorinostat resulted in an increase in acetylated α-tubulin and histone 3. Similar results were obtained in 4 T1-Luc mouse breast cancer cells (Supporting Information Fig. 1B). The selective inhibition of HDAC6 by Tubathian A at 1 μM was conserved throughout cancer cell lines from multiple types of solid tumours, comparable to the effect of Tubastatin A (Supporting Information Fig. 1C). When treating SK-OV-3-Luc cells with this low con- centration followed by removal of the HDAC6i, control levels of K-40 acetylated α-tubulin were reached 1 h after wash out, indicating the reversibility of the interaction. A similar reversible effect was obtained when using Tubastatin A (Fig. 1c). K- 40 acetylation of α-tubulin increased within 30 min after administration of both compounds and was maintained dur- ing the course of the experiment (Fig. 1d), indicating a persistent effect in their continuous presence. High concentrations of HDAC6i likely induce off-target effects To elucidate if the observed effects are caused by increased HDAC6 inhibition at high concentrations or by off-target effects, the HDACi were tested on a CRISPR/Cas9 mediated HDAC6 knockout cell line, HAP1-KO. Immunoblotting results confirmed the complete lack of HDAC6 protein expression, which resulted in a strong elevated level of K-40 acetylated α-tubulin in HAP1-KO cells compared to the parental strain (Fig. 3a). When treating HAP1-KO cells with the HDACi, no increase in acetylated α-tubulin occurred, con- firming the lack of HDAC6 expression (Fig. 3b and Support- ing Information Fig. 3A). When using the selective HDAC6i Tubathian A, Tubastatin A and Tubacin at 10 μM, increased levels of K-9 acetylated histone 3 were observed, indicating a loss of selectivity at high concentrations. In accordance with the effects observed in SK-OV-3-Luc and 4 T1-Luc cells, the HAP1 parental cells demonstrated identical responses with regards to acetylated α-tubulin and histone 3 upon treatment with Tubathian A (Supporting Information Fig. 3B). Vorinostat increased the acetylated histone 3 level at concentrations as low as 1 μM (Fig. 3b). When analysing the cell confluency of HAP1-KO cells treated with the different HDACi, no effects were observed using HDAC6-selective concentrations, as expected due to the lack of the target enzyme (Fig. 3c and Supporting Information Fig. 3C). High concentrations of even the most selective HDAC6i resulted in a severe stagnation and reduction of cell confluency. Furthermore, when analysing the metabolic activity of HAP1-KO and HAP1 cells after treat- ment with both high and low concentrations of HDACi, no difference in response could be observed between the two cell lines, indicating lack of HDAC6 specificity (Fig. 3d). Tubathian A shows in vivo anti-tumour activity only when exhibiting non-selective properties To translate our in vitro findings to an in vivo setting, a peri- toneal metastasis model was used, in which SK-OV-3-Luc human ovarian cancer cells were intraperitoneally (i.p.) injected in Swiss nude mice. Daily i.p. treatment with control (vehicle, N = 7) or treatment (5 and 20 mg/kg Tubathian A; N = 7 and 9, respectively) resulted in a significant reduction in bioluminescence at five weeks post inoculation only in the 20 mg/kg Tubathian A condition (Fig. 4a). Consequently, the survival of the latter mice increased significantly (Fig. 4b). No effect was observed on body weight in any of the treatment groups (Supporting Information Fig. 4A). Analysis of the tumour tissue showed an increased level of K-40 acetylated α-tubulin in both treatment groups 1 h after injection (Figs. 4c and 4d). However, only the 20 mg/kg Tubathian A dose increased the level of K-9 acetylated histone 3 in the tumour tissue, indicating the loss of HDAC6 selectivity. Hematoxylin and eosin (H&E) staining of peritoneal metastasis showed epi- thelial nests intermingled with cancer-associated fibroblasts in all conditions. Infiltration into the fat tissue was reduced in 20 mg/kg Tubathian A condition (Fig. 4e). The Ki67 nuclear proliferation marker did not significantly change between control and HDAC6 specific inhibition (5 mg/kg Tubathian A), however again a significant decrease was observed when high concentrations (20 mg/kg Tubathian A) were used (Fig. 4f ). To extend our observations in another in vivo model and to explore possible contributions of immune-response changes induced by HDAC6 inhibition, we used an immune-compe- tent, metastatic mouse xenograft model. 4 T1-Luc murine breast cancer cells were orthotopically injected in the mam- mary fat pad of BALB/c mice, which were randomized in three groups. Mice were i.p. treated daily with control (vehicle, N = 6) or treatment (5 and 20 mg/kg Tubathian A, N = 6 and 7 respectively) and sacrificed six weeks post inoculation. No effect was observed on the weight of the primary tumour, nor on its ex vivo bioluminescence signal (Figs. 5a and 5b). The percentage of mice bearing lung metastasis as well as the ex vivo bioluminescence signal of the lung metastasis, were not influenced by the treatment (Figs. 5c and 5d). Analysis of the primary tumour tissue showed an increased level of K-40 acetylated α-tubulin upon treatment with both 5 and 20 mg/kg Tubathian A, without influencing the level of K-9 acetylated histone 3, pointing to selective HDAC6 inhibition at both concentrations in this model (Figs. 5e and 5f ). H&E staining indicated mildly differentiated primary tumours inva- sive into the muscular tissue and associated with a lympho- cytic reaction and lungs showed metastatic nodules (Fig. 5g). No differences in necrotic area, degree of muscular invasion, lymphocyte infiltration nor size of metastatic nodules were observed between all treatment conditions. Discussion HDAC6 has been promoted as a therapeutic target in cancer research.11 Substantial effort has been devoted to the develop- ment of selective HDAC6i and their use as anti-cancer agents. Tubathian A, a sulphur analogue of Tubastatin A, showed excellent HDAC6 inhibitory capacities in cancer cell lines from multiple tumour types, elevating the levels of K-40 acetylated tubulin already at low concentrations. Treatment with Tubathian A, when used in concentrations up to 1 μM, only minimally increased the levels of K-9 acetylated histone 3 demonstrating good selectivity towards HDAC6 relative to class I HDACs. Higher concentrations did result in strong inhibition of nuclear HDAC enzymes. When treating with Tubastatin A or Tubacin, similar biochemical effects were observed. This non-selective inhibition of class I HDACs by the HDAC6i Tubastatin A at high concentrations has previously been reported.15 As expected based on enzymatic selectivity data, both Tubathian A and Tubastatin A showed a better selectiv- ity profile compared to Ricolinostat, an HDAC6i currently undergoing clinical trials.23 Given the anti-cancer properties attributed to HDAC6 inhibition, Tubathian A was tested for its impact on growth and migratory and invasive properties in established cell lines from multiple tumour types. The compound, as well as Tubas- tatin A and Tubacin, all failed to induce any anti-cancer prop- erties when used in an HDAC6-selective manner. Only high concentrations of these compounds, resulting in non-selective HDAC inhibition, reduced growth and, in some cases, migra- tion/invasion of cancer cells. Furthermore, Tubathian A failed to reduce cancer progression in both an in vivo peritoneal metastasis and breast cancer metastasis model, despite show- ing selective HDAC6 inhibition. When used in a high concen- tration, Tubathian A induced non-selective HDAC inhibition within SK-OV-3-Luc tumours and resulted in supressed peri- toneal metastasis and prolonged survival of the mice. The most frequently reported HDAC6i are Tubastatin A, Tubacin and Ricolinostat and their use as anti-cancer therapeutics is summarized in Supporting Information Table 1. Interestingly, these three compounds most often display anti-cancer proper- ties when used in micromolar ranges in an in vitro setting and high in vivo concentrations (>10 mg/kg), undoubtedly exhibiting inhibition of other HDAC enzymes besides HDAC6, in agreement with our results. Unfortunately, in most of the papers consequent control on the level of K-9 acetylation of histone 3 was not investigated. These results raise doubt whether the observed effects at these concentra- tions are indeed HDAC6 related, or whether they occur because of off-target effects. We compared the response of HAP1 cells with their HDAC6 knockout isogenic line and both showed the same functional response to all used HDACi, indicating that HDAC6 inhibition does not play a role in the observed effects. Translating these results to literature data suggests that many reported anti-cancer properties of HDAC6i, are in fact most probably unrelated to selective HDAC6 inhibition and can be attributed to off-target effects.
Next to these well-known HDAC6i, many more com- pounds have been synthesized and tested, mostly for their ability to reduce cancer cell growth, indicating the potential of HDAC6i for cancer treatment. The selectivity index (IC50 of HDAC1/HDAC6) and active concentrations of all these inhib- itors are reported in Supporting Information Table 2, illustrat- ing that most of these inhibitors display activity only when used in micromolar concentrations. Since they in general pos- sess inferior selectivity profiles compared to Tubathian A and Tubastatin A, and in most cases even to Tubacin, off-target effects cannot be excluded. Remarkably, almost all inhibitors that do display nanomolar anti-cancer activities, possess IC50 values against HDAC1 in the same range.
Interestingly, a minority of papers still reports the use of HDAC6i at low nanomolar concentrations. Meng et al. showed that the use of Tubastatin A in a concentration as low as 4 nM reduces glioblastoma cell migration.26 Tubastatin A at 500 nM resulted in decreased cell growth of glioblastoma cells and reduced migration of oesophageal squamous cell car- cinoma cells.27,28 Additionally, Ricolinostat reduced glioblas- toma cell growth in a 10 nM concentration, indicating a potential use of HDAC6 inhibitors for these cancer types.29 Tubacin has been shown to inhibit Burkitts lymphoma cell migration in a concentration of 1 μM. Remarkably, the same study showed that Niltubacin, a catalytically inactive Tubacin analogue, showed similar results, indicating the possibility of participating Tubacin properties that are unrelated to its abil- ity to inhibit HDAC6 deacetylase activity, even at low concen- trations.30 This observation was reinforced by the finding that Tubacin interfered with sphingolipid biosynthesis and lym- phocyte chemotaxis independently of its HDAC6 deacetylase activity.31,32 We suggest to take into account this possibility when interpreting the anti-migratory properties of Tubacin at low concentrations regarding breast cancer cells and neuro- blastoma in literature.33,34 Furthermore, these effects could contribute to the growth-inhibitory and anti-migratory prop- erties of Tubacin in our study on SK-OV-3-Luc 4 T1-Luc and HAP1 cells. Valente et al. report a tert-butylcarbamate based inhibitor of neuroblastoma cell growth at an IC50 of 0.4–0.6 μM.
Other cancer cell lines were tested, which displayed less sensitivity (e.g. SK-OV-3 cells, IC50 = 40 μM) pointing towards potentially high sensitivity of the neuroblastoma cells to HDAC6 inhibition.35 Compound C1A, reported by Kaliszc- zak et al. also displays excellent growth inhibition of neuro- blastoma cells.36 Remarkably, the mean growth inhibitory effect on some of the tested cell lines (e.g. Kelly: 0.18 μM) was far below the compound’s enzymatic IC50 value against HDAC6 (0.479 μM), which renders the involvement of HDAC6 inhibition questionable.
In addition to the direct effect of HDAC6 inhibition on cancer cells, it has been reported to interfere with the immune regulation as well. Nevertheless, we did not observe an effect on tumour progression and metastasis formation in an immune-competent 4 T1-Luc breast cancer model using Tubathian A. Mixed literature results make it hard to predict any immune-related responses to selective HDAC6 inhibition, further complicated by the possible off-target effects caused by high concentrations. Tubastatin A (3 μM) and Nexturastat A (3 μM) increased MHC class I and tumour antigen expression and downregulate PD-L1 expression in melanoma cells.37,38 In contrast, the inhibition of HDAC6 with Tubastatin A (2 μM) in colorectal cancer cells increased PD-L1 expression, while lower concentrations do not induce an effect.39 Woods et al. showed that selective HDAC6 inhibition (Nexturastat A (1 μM) and Ricolinosat (0.5 μM)) did not result in altered expression of PD-L1 in melanoma cells, while the use of non-selective inhibitors increased PD-L1 expression.40 Further- more HDAC6 inhibition by Tubacin (0.1 μM) and Tubastatin A (1 μM) increases the suppressive function of Foxp3 Tregulatory cells over Teffector cells.41,42
Overall, our results strongly indicate a gap between the clear biochemical changes caused by HDAC6i, namely the increased acetylation of HDAC6 substrates, and consequent functional effects. Functional insights into the roles of the acetylation level of these substrates, have mostly been obtained by interfering with the activity or expression of HDAC6.9,14,43–45 Especially α-tubulin K-40 acetylation has been promoted as an important regulator of cell motility and adhesion, given the negative effect of HDAC6 downregulation on these processes.9,43
More recently, the manipulation of α-tubulin K-40 acetyltransferase (αTat1) provided new insights in the possible role of acetylated α-tubulin. Genetic abrogation of αTat1 results in loss of contact inhibition of proliferation as well as reduction of focal adhesions in fibroblasts, while the invasion of breast cancer cells in collagen matrix is negatively influenced by decreasing αTat1 expres- sion.46,47 Since both HDAC6 and αTat1 have multiple sub- strates and/or deacetylase-independent functions, it is very difficult to draw straightforward conclusions from these results concerning the functional impact of α-tubulin K-40 (de)acetylation. It was already shown that HDAC6 KO mice develop normally and display almost no altered phenotypes, indicating the lack of impact of the increased acetylation level of HDAC6 substrates on normal mammalian development, and our data indicate that cancer cells as well are remarkably tolerant to very large changes in the degree of side chain lysine acetylation of HDAC6 substrates, including tubulin acetylation.13
Besides the inhibition of the deacetylase capacity of HDAC6, knockdown of the expression of this enzyme using siRNA or shRNA, has been able to affect growth, migration and invasion of a variety of cancer cell lines, indicating a clear role for HDAC6 in cancer progression (Supporting Informa- tion Table 4). Since our results, and in extension the majority of literature data, do not point to an involvement of its deace- tylase function in controlling these processes, we consider it likely to contribute through deacetylase-independent proper- ties. The divergence between results observed using siRNA and CRISPR/Cas9 KO versus those observed using pharmaco- logical methods is typically attributed to protein–protein interactions. An enzyme treated with an inhibitor may still act as a scaffold for protein–protein interactions, which would be disrupted by a knockdown or knockout strategy.48 Still, the introduction of a catalytically inactive mutant HDAC6 in SK-OV-3 cells knocked down for HDAC6 by shRNA, showed the need for deacetylase active HDAC6 for cancer cell growth in vitro and in vivo.49 Although the impact of these genetic manipulations was not evaluated on histone acetylation nor confirmed by HDAC6 specific pharmacologic inhibition, the present study failed to substantiate these results by pharmaco- logical inhibition of the deacetylase activity of HDAC6 by a variety of HDAC6i in the same cell line.
Our biological evaluation of HDAC6i Tubathian A, Tubastatin A, Tubacin and Ricolinostat, combined with a crit- ical literature review, questions the usefulness of selective HDAC6i as single agent cancer therapeutics. Nonetheless, it is worth mentioning that many research efforts are being devoted to the use of HDAC6i in combination with standard chemotherapies. Again, many of the literature reports need to be carefully interpreted due to the use of high molarities and non-selective compounds, such as ACY-241.50 Nonetheless, promising results have been obtained combining selective HDAC6i at low concentrations with Temozolomide, an alky- lating agent used for the treatment of glioblastoma multi- forme, or with Cisplatin in non-small cell lung cancer cells.27,51–53 One of the most studied combinatory therapies concerning HDAC6i, is their use with proteasome inhibitors, such as Bortezomib and Carfilzomib, for the treatment of multiple myeloma. In most cases Ricolinostat was used, but with concentrations ranging from 1 to 4 μM, this approach is likely to result in off-target effects on other HDACs.23,54,55 In fact, it was shown that Ricolinostat, at a clinically relevant dose of 160 mg (twice daily), in combination with Bortezomib in patients, elevated the level of acetylated histones in periph- eral blood lymphocytes taken 1 h after the first Ricolinostat dose.56 Synergistic effects combining Tubacin or Tubastatin A and Bortezomib/Carfilzomib when used against multiple mye- loma, breast cancer and ovarian cancer, were only present at high molarities.55,57–59 All together, these results call for cau- tion when claiming synergy between HDAC6i and proteasome inhibitors.
As a final remark, recent literature pointed towards the dependency on HDAC6 activity in ovarian cancers with func- tional p53 combined with ARID1A mutation.60 Therefore, the lack in activity we observed in SK-OV-3-Luc ovarian cancer cells, might at least be partially explained by the fact that these cells are ARID1A mutated but do not express p53.61 Gener- ally, this might indicate that the effect of HDAC6 inhibition could be strongly dependent on cancer subtype.
In conclusion, the inhibition of HDAC6 by Tubathian A and other HDAC6i did not result in any anti-cancer proper- ties. Only when using non-selective conditions, i.e. high con- centrations or less selective inhibitors, effects on cancer cell growth, migration and/or invasion, and in vivo tumour pro- gression were observed. We strongly emphasize that HDAC6 specificity should be evaluated when testing HDAC6i and question the general usefulness of ACY-1215 selective HDAC6i as single treatment cancer therapeutics.