HDAC inhibitors have been reported to produce antidepressant and pro-cognitive effects in animal models, however, poor brain bioavailability or lack of isoform selectivity of current probes has limited our understanding of their mode of action. We report the characterization of novel pyrimidine hydroxyl amide small molecule inhibitors of HDAC6, brain bioavailable upon systemic administration. We show that two compounds in this family, ACY-738 and ACY-775, inhibit HDAC6 with low nanomolar potency and a selectivity of 60- to 1500-fold over class I HDACs. In contrast to tubastatin A, a reference HDAC6 inhibitor with similar potency and peripheral activity, but more limited brain bioavailability, ACY-738 and ACY-775 induce dramatic increases in α-tubulin acetylation in brain and stimulate mouse exploratory behaviors in novel, but not familiar environments. Interestingly, despite a lack of detectable effect on histone acetylation, we show that ACY-738 and ACY-775 share the antidepressant-like properties of other HDAC inhibitors, such as SAHA and MS-275, in the tail suspension test and social defeat paradigm. These effects of ACY-738 and ACY-775 are directly attributable to the inhibition of HDAC6 expressed centrally, as they are fully abrogated in mice with a neural-specific loss of function of HDAC6. Furthermore, administered in combination, a behaviorally inactive dose of ACY-738 markedly potentiates the anti-immobility activity of a subactive dose of the selective serotonin reuptake inhibitor citalopram. Our results validate new isoform-selective probes for in vivo pharmacological studies of HDAC6 in the CNS and reinforce the viability of this HDAC isoform as a potential target for antidepressant development.
Originally named in relation to their role in chromatin regulation, HDACs comprise a family of 11 structurally-related zinc-binding hydrolases, which catalyze the removal of acetyl moeties from lysine residues on proteins (Bradner et al, 2010). Deacetylation of N-terminal histone tails is the predominantly studied function of HDACs. However, recent proteomics studies have also identified a wide variety of non-histone substrates for HDACs, including transcription factors, cytoskeletal proteins, metabolic enzymes, and chaperones (Choudhary et al, 2009). These non-histone substrates are likely to contribute at least in part to the pharmacological activity of HDAC inhibitors reported in therapeutic areas like cancer, autoimmune diseases, and neuropsychiatric disorders. Early evidence implicates HDACs, in the pathophysiology and treatment of mood disorders. HDACs 2, 4, 5, 6, and 8 mRNAs levels are altered in blood cells or post-mortem brain of patients diagnosed with mood disorders (Covington et al, 2009; Hobara et al, 2010). Sodium valproate, a drug used in the clinical treatment of mood disorders has HDAC inhibitory properties (Phiel et al, 2001). Furthermore, expression levels of various HDACs are modulated by antidepressants and mood stabilizers in neuronal and non-neuronal tissue culture systems (Chen et al, 2010; Mao et al, 2011). Finally, several HDAC inhibitors, including SAHA, sodium butyrate, and MS-275, and the acetylating agent L-acetylcarnitine have been shown to promote antidepressant-like behavioral responses in rodent models (Covington et al, 2009; Golden et al, 2013; Nasca et al, 2013; Russo and Charney, 2013). The relative contribution of histone vs non-histone HDAC substrates to these antidepressant-like effects is not known.
Recently, along with others, we identified a role for HDAC6 in models of stress adaptation and showed that HDAC6 loss of function results in an antidepressant-like phenotype (Espallergues et al, 2012; Fukada et al, 2012; Lee et al, 2012). These results point to a role of non-histone substrates in the antidepressant-like activity of pan-HDAC inhibitors. HDAC6 is a structurally and functionally unique class IIb isoform localized almost exclusively to the cytoplasm (Verdel et al, 2000) that does not appear to deacetylate histones in vivo (Zhang et al, 2003). The most abundant and best-characterized substrate of HDAC6 is α-tubulin (Hubbert et al, 2002), with a number of other HDAC6 substrates identified, including Hsp90 (Kovacs et al, 2005), cortactin (Zhang et al, 2007), and beta-catenin (Li et al, 2008).
Although first-generation hydroxamates such as TSA or SAHA indiscriminately inhibit HDAC6 and class I isoforms (HDACs 1, 2, and 3; Bertrand, 2010), small molecules developed more recently such as tubacin (Haggarty et al, 2003) and tubastatin A (Butler et al, 2010) have been shown to afford more selective inhibition HDAC6. These probes have been instrumental in validating HDAC6 as a therapeutic target in models of neurodegenerative disease (d'Ydewalle et al, 2011).
Recent studies combining the use of these pharmacological probes and genetic approaches have provided indications that the physiological impact of selective HDAC6 pharmacological inhibition or loss of function in neurons diverges drastically from that of other HDACs (Rivieccio et al, 2009). In contrast to class I isoforms, the loss of function of HDAC6 does not produce toxicity, apoptosis, or major neurodevelopmental defects in rodents (Govindarajan et al, 2012; Haberland et al, 2009; Zhang et al, 2008).
Furthermore, early clinical trials with novel HDAC6 inhibitors suggest that, despite their significant biological activity, these experimental drug candidates appear to largely circumvent undesirable side-effects classically reported with broad-acting or class I-selective inhibitors (Raje et al, 2013). However, because pharmacokinetic properties of current HDAC6-selective probes are not optimized for in vivo CNS studies, it remains unclear whether systemic pharmacological inhibition of HDAC6 can produce antidepressant-like effects in animal models.
Here we report the behavioral activity of two novel small molecule inhibitors of HDAC6 with brain bioavailability, low nanomolar potency, and selectivity over class I HDACs. We provide evidence that systemic administration of these probes induces dramatic increases in brain α-tubulin acetylation, without concurrently altering the acetylation of histones. Further, using mice with conditional loss of function of HDAC6 we show that, through their selective inhibitory activity on HDAC6 expressed in the CNS, these probes replicate the behavioral effects of acutely and chronically administered reference antidepressants.
MATERIALS AND METHODS
Male NIH Swiss mice ordered from Harlan Laboratories (Frederick, MD) at 6 weeks of age (18–20 g) were group housed and acclimated for 1 week before testing. For experiments involving HDAC6 loss of function, we used a previously described flHDAC6nestin-cre line (Espallergues et al, 2012) with a neural cell-specific KO of HDAC6. Mice were backcrossed on C57BL/6J background for over 10 generations. Animals were allowed food ad libitum and were tested at 6–10 weeks of age during the light phase of 12-h cycle (0700–1900 hours). All protocols were approved by the University of Pennsylvania Institutional Animal Care and Use Committee.
ACY-738 (N-hydroxy-2-(1-phenylcycloproylamino)pyrimidine-5-carboxamide) and ACY-775 (2-((1-(3-fluorophenyl)cyclohexyl)amino)-N-hydroxypyrimidine-5-carboxamide) were synthesized by ChemPartner (Shanghai, China) and obtained from Acetylon Pharmaceuticals. Other drugs used: tubastatin A, a gift from A Kozikowski (University of Illinois, Chicago); trichostatin A (TSA, Sigma-Aldrich, St Louis, MO); fluoxetine-HCl (AK Scientific, Union City, CA); citalopram-HBr (AK Scientific); chlordiazepoxide-HCl (CDP, Sigma-Aldrich). For in vitro assays, drugs were dissolved and diluted in assay buffer (50 mM HEPES, pH 7.4, 100 mM KCl, 0.001% Tween-20, 0.05% BSA, and 20 μM tris) to sixfold the final concentration. For pharmacokinetic analyses, drugs were dissolved in 10% DMAC, 10% solutol HS15, and 80% saline at a dose of 5 mg/kg. For in vivo behavioral and biochemical assays, all drugs were dissolved in DMSO and diluted to a final concentration of 0.75% in 0.9% saline. Owing to limited solubility, the 50 mg/kg dose of ACY-775 was in a suspension state when injected. All intraperitoneal (i.p.) injections were administered at a volume of 10 ml/kg.
HDAC Enzymatic Assays
Potency of each compound was determined using recombinant enzymes and a kinetic method described previously (Bradner et al, 2010). Briefly, N terminal GST-tagged HDAC enzymes (BPS Biosciences, San Diego, CA) expressed in a baculoviral system and purified by affinity chromatography and gel filtration were diluted to 1.5-fold final concentration in assay buffer and pre-incubated with test compounds for 10 min before the addition of substrate, FTS (HDAC1, HDAC2, HDAC3, and HDAC6) or MAZ-1675 (HDAC4, HDAC5, HDAC7, HDAC8, and HDAC9). For studies with HDAC6 N- and C-terminal catalytic mutants, point-mutations H216A and H611A were used, respectively (Haggarty et al, 2003). The amount of substrate for each enzyme was equal to the Michaelis constant (Km), as determined by a titration curve. The enzymatic reaction was monitored over 30 min for release of 7-amino-4-methoxy-coumarin after deacetylation of the lysine side chain in the peptide substrate and the linear rate of the reaction was calculated.
For PK analyses tubastatin A, ACY-738 and ACY-775 were dosed cassettely and measured in plasma and brain by liquid chromatography/tandem mass spectrometry. Plasma and brain were collected from mice killed 5, 15, 30, and 60 min after treatment. Drugs were extracted from brain and plasma by protein precipitation using 50 : 50 acetonitrile:methanol and analyzed using a HPLC/MS/MS method using electrospray ionization in positive mode as previously reported (Santo et al, 2012). The lower limit of quantification for all compounds was 3 ng/ml. PK parameters (concentrations, T1/2, AUCBrain/AUCPlasma) were estimated by non-compartmental model using WinNonlin (Pharsight, Sunnyvale, CA).
Undifferentiated RN46A-B14 cells, a line of immortalized rat raphe neuronal precursors, were grown as previously described (White et al, 1994). They were treated with 2.5 μM ACY-738, ACY-775, tubastatin A, or 0.6 μM TSA, or vehicle (0.1% DMSO) for 4 h. Samples were processed using Epigentek (Farmingdale, NY) histone extraction kit and quantified using protein assay (Bio-Rad, Hercules, CA).
Brains were harvested after cervical dislocation and decapitation, flash frozen, and stored at −80 °C until use. Needle punches from cortex (bregma 2.22 mm), hippocampus (bregma −1.34 mm), dorsal raphe nucleus (DRN; bregma −4.36), and cerebellum (bregma −5.34) were collected. Extracts from whole brains and punches were harvested and quantified as above.
Standard immunoblotting procedures were used. Antibodies: total α-tubulin (1 : 2000, Rockland Immunochemicals, Gilbertsville, PA), acetyl α-tubulin (1 : 20 000, Sigma-Aldrich), acetyl lysine 9 histone H3 (1 : 500, Abcam, Cambridge, MA), and total H3 (1 : 1000, Abcam). Blots were imaged using LiCOR (Lincoln, NE) Odyssey system and quantified using ImageJ (NIH, Bethesda, MD).
Chromatin Immunoprecipitation (ChIP)
Mice were subchronically treated with ACY-738 (5 mg/kg), sodium butyrate (1.2 g/kg), or vehicle. Chromatin extraction and ChIP was performed as previously reported (Zhang et al, 2005). Hippocampus was harvested 30 min after last injection and immediately crosslinked in 1.1% formaldehyde/PBS for 10 min. Immunoprecipitation with IgG or anti-AcH3K9 (Abcam) antibodies on 5 μg of chromatin was performed as reported previously (Kennedy et al, 2013). QPCR was completed on precipitated chromatin for the promoter sequences of interest (BDNF IV and cFOS; Guan et al, 2009). Promoter enrichment was calculated using primers flanking 18S rRNA loci as a control for both genomic, input DNA, and precipitated DNA. Primer sequences are available on request.
Mice were injected with ACY-738 or ACY-775 at 5, 10, or 50 mg/kg or vehicle and allowed to explore. Activity was recorded using the Photobeam Activity System (PAS, San Diego Instruments, San Diego, CA).
Home Cage Activity
Activity was recorded using an infrared beam, home cage activity system from Med Associates (St Albans, VT) immediately following i.p. injection of 50 mg/kg ACY-738 or vehicle.
Fifteen minutes after injection, mice were placed in a standard mouse cage with 6 cm bedding and 20 (4 × 5) clear glass marbles. After 15 min, an investigator blinded to the treatment counted the number of marbles buried to at least three-fourths.
Elevated Plus Maze (EPM)
Fifteen minutes after injection, mice were tracked in the EPM (Lister, 1987) using TopScan software (Clever Sys, Reston, VA).
Novelty-Induced Hypophagia (NIH)
NIH was performed as previously described (Merali et al, 2003) using peanut butter chips as a palatable food. Mice were tested 30 min after injection for latency to eat in a novel environment including an unfamiliar scent and bright (600 lux) lighting.
Tail Suspension Test (TST)
Mice were tested for immobility in the TST using a rig from Med-Associates. At 30 min or 2 h after i.p. injection of ACY-738 (5, 50 mg/kg), ACY-775 (5, 50 mg/kg), and citalopram (0.5, 2, 20 mg/kg), a combination of the previous, or vehicle, mice were attached to the test rig and time immobile over 6 min was recorded.
Chronic Social Defeat (CSD) and Social Interaction Testing
Social defeat was conducted as previously described (Berton et al, 2006; Golden et al, 2011). For stratification of resilient/vulnerable sub-population after social defeat exposure, we used the score of the lowest-interacting undefeated control mouse as a cutoff. Any defeat-exposed mouse that spent over 51.8 s in the interaction zone with target present was categorized as resilient.
Open-field data were analyzed using a two-way ANOVA with repeated factor, followed by Bonferroni post-tests. For other data sets, one-way ANOVA followed by Neumann–Keuls post-hoc test or t-test were done as appropriate. Statistical analyses were conducted using GraphPad Prism 5.0a (Graph Pad Software, San Diego, CA). Drug effects on the proportion of resilient mice after social defeat were computed using z-test in Excel (Microsoft, Redmond, WA).
ACY-738 and ACY-775 demonstrate inhibitory activity against recombinant HDAC6 with IC50 values of 1.7 and 7.5 nM, respectively, with respective average selectivity over class I HDACs being 100- and 700-fold (Table 1, Figures 1b–d). For comparison, the reference HDAC6 inhibitor tubastatin A showed an IC50 for HDAC6 of 18 nM and displayed an average 200-fold selectivity over class I HDACs (Figure 1a). ACY-775 had minimal activity against all other class II HDAC isoforms (IC50>1 μM). Activity of ACY-738 and ACY-775 in in vitro assays was fully explained by inhibition of the C terminal catalytic domain of HDAC6, as deacetylase activity of a point-mutant with inactive N-terminal domain (H216A) but intact C-terminal domain was comparable to HDAC6 WT and fully inhibited by ACY-738. In contrast, the HDAC6 C-terminal catalytic mutant was almost completely devoid of deacetylase activity (Supplementary Figure S1).
In vivo biodistribution profiles of ACY-738, ACY-775, and tubastatin A were examined after acute dosing at 5 or 50 mg/kg over 2 h. At t=30 min after acute 50 mg/kg injection, respective plasma levels of ACY-738 and ACY-775 were 515 ng/ml (1.9 μM) and 1359 ng/ml (4.1 μM). Elimination from plasma was rapid, with plasmatic half-life of 12 min and concentration below 10 ng/ml after 2 h. Nevertheless, areas under concentration time curves for brain and plasma (AUCBrain/AUCPlasma) calculated over 2 h for both ACY-738 and ACY-775 led to ratios >1 (Table 1). In comparison, tubastatin A exhibited a longer plasmatic half-life of 2 h, but a more limited brain penetration, with AUCBrain/AUCPlasma ratio of 0.18. Taken together, these results suggest that despite their short half-life, ACY-738 and ACY-775 rapidly distribute to the brain leading to a total drug exposure in CNS comparable to that of peripheral tissues.
In contrast to the class I isoforms HDAC2 and 3, which are expressed highly and ubiquitously in the mouse brain, HDAC6 mRNA expression appears restricted to a small number of brain areas, with highest signal observed in DRN (Figure 1f). In RN46A-B14 cells treated with TSA (0.6 μM), tubastatin A (2.5 μM), ACY-738 (2.5 μM), or ACY-775 (2.5 μM), western blot analyses at 4 h after treatment revealed increases in the acetylated (lysine 40) fraction of α-tubulin (F4, 9=48.69, P<0.0001). No changes in total α-tubulin expression were detected (Figure 2a). Although TSA, led to a 94% increase in acetylation of histone H3 at lysine 9 (F4, 5=22.21, P=0.0022), no significant change in histone H3K9 acetylation was observed following treatment with ACY-738, ACY-775, or tubastatin A (Figure 2a).
Although tubastatin A resulted in an increase in α-tubulin acetylation of about 400% in tissue culture (Figure 2a), and produced 268% increased α-tubulin acetylation in heart (P<0.05) upon single systemic administration in vivo (10 mg/kg, 2.7 mM), it did not significantly change levels of α-tubulin acetylation in whole-brain lysates (Figure 2b). In contrast, ACY-738 (5 mg/kg, 1.9 mM) and ACY-775 (50 mg/kg, 15 mM), both led to significant increase in α-tubulin acetylation in whole-brain lysates. Changes were significant at 30 min (F3, 13=163.4, P<0.0001), 1 h (F3, 13=163.9, P<0.0001), and 4 h (F3, 14=4.703, P=0.0179) after administration.
When ACY-738 (5 mg/kg) or ACY-775 (50 mg/kg) were administered repeatedly in wild-type mice at 24 h, 4 h, and 30 min before killing, significant increases in α-tubulin acetylation were observed in all tested brain regions (Figure 2c): cortex, F2, 7=582.5, P<0.0001; hippocampus, F2, 7=260.4, P<0.0001; DRN, F2, 7=54.00, P<0.0001; and cerebellum, F2, 7=136.2, P<0.0001. In contrast, an identical treatment regimen in KO mice did not produce increases in α-tubulin acetylation over baseline levels (Figure 2c). Of note, as reported previously (Espallergues et al, 2012), these mutant mice have markedly enhanced baseline acetylated α-tubulin (1.98±0.2 acetyl/total) when compared with their WT littermates (0.5±0.06; P<0.0001).
As reported above in cell culture, treatment with ACY-738 and ACY-775 in vivo did not significantly alter acetylation of histone 3 at lysine 9 in western blot. In addition, ChIP revealed no concurrent changes in histone H3K9 acetylation enrichment at promoter DNA within the hippocampus following repeated i.p. treatment with 5 mg/kg ACY-738. This is contrast to treatment with a behaviorally active dose of sodium butyrate (1.2 g/kg), a class I histone deacetylase inhibitor, which led to increases in enrichment of acetylated H3K9 at the promoter of activity-induced neuronal genes, namely cFOS and BDNF promoter 4 (Figure 2d and e).
Common antidepressants have been shown to increase ambulation in a novel environment (Brocco et al, 2002). We tested the effects of acute administration of ACY-738, ACY-775, and tubastatin A on exploration in a novel open-field arena. When injected after 1 h of habituation, tubastatin A did not change exploratory activity at 10 mg/kg. However, we observed rapid, dose-dependent hyperlocomotor effects after treatment with 50 mg/kg ACY-738 (F4, 132=84.63, P<0.0001; Figures 3a and c) or ACY775 (F4, 174=7.265, P<0.0001; Figure 3b and c). There was a significant effect of time (ACY-738: F5, 132=84.63, P<0.0001; ACY-775: F5, 174=28.71, P<0.001), with effect peaking in the first hour after injection and returning to baseline by 2 h, consistent with brain biodistribution profiles. The same doses of ACY-775 or ACY-738 did not significantly alter open-field exploration in mice with neural cell-selective KO HDAC6 (Figures 3a–c). Distance in the center of the open field trended toward an increase, with ACY-738 (602±96 beam breaks) and ACY-775 (419±73 beam breaks) treated mice traveling a greater distance in the center than those treated with vehicle (177±71 beam breaks). Furthermore, a 50 mg/kg dose of ACY-738 failed to produce an enhancement of locomotor activity in WT mice tested in a home cage environment (veh=4273±425 beam breaks; ACY-738=3468±711 beam breaks; F1, 232=0.9447, P=0.36), suggesting that the effect reflects a disinhibition of exploratory behavior under neophobic condition, rather than a nonspecific elevation of motor function.
In the marble-burying task, a test sensitive to both anxiolytics (Broekkamp et al, 1986) and antidepressants (Albelda and Joel, 2012), treatment with ACY-738 and ACY-775 significantly reduced the number of marbles buried, to a similar degree as a non-sedating (10 mg/kg) dose of the common anxiolytic CDP (Figure 3d; F3, 50=3.699, P=0.0176). In addition, ACY-738 given acutely at a dose of 50 mg/kg significantly decreased latency to eat in the NIH test (P<0.05), although this effect was not as potent as that of CDP (P<0.01; Figure 3e). In the EPM, a test sensitive to anxiolytics but not antidepressants, neither ACY-738 nor ACY-775 produced behavioral changes, whereas CDP markedly increased the time in the open arms of the maze (Figure 3f; F3, 16=7.498, P=0.0024).
In the TST (Figure 4a), we observed a significant effect of drug/dose (F12, 140=9.368, P<0.0001). Significant reductions in immobility were detected at 50 mg/kg with either ACY-738 (P<0.001) or ACY-775 (P<0.01). No significant effects of either drug were seen after a 5 mg/kg dose. Administration of the SSRI citalopram led to significant decreases in time immobile at 2 mg/kg (P<0.001) and 20 mg/kg (P<0.001), but not at 0.5 mg/kg. At 2 h after injection, treatment with 20 mg/kg citalopram (P<0.001) or 50 mg/kg ACY-738 (P<0.01) still had significant activity.
To test whether anti-immobility effect of citalopram and ACY-738 are dependent on intact HDAC6 protein levels, we administered citalopram to mice with neural KO of HDAC6 or their WT littermates bred on a C57BL/6J background (Figure 4b). In line with previous reports comparing C57BL/6J with NIH Swiss mice in TST (Lucki et al, 2001), we found that, while anti-immobility effects were still significant, C57BL/6J WT had higher baseline immobility (P<0.0001) and were less responsive to both citalopram (P<0.0001) and ACY-738 (P<0.01) than NIH Swiss mice. In contrast, the anti-immobility effect of citalopram was significantly greater in HDAC6 KO mice than WT littermates (F4, 47=7.902, P<0.0001), in line with a previous report showing that the activity of another SSRI, fluoxetine, is amplified in HDAC6 KOs (Fukada et al, 2012). However, when anti-immobility activity of ACY-738 was tested in HDAC6 KO mice, no significant effect was observed, indicating that intact HDAC6 protein is required for ACY-738 action in TST.
To further examine the interactions between SSRI and HDAC6, we tested the effects of a combined administration of citalopram and HDAC6 inhibitors in NIH Swiss mice (Figure 4a). We found significant anti-immobility effects of a subactive dose of citalopram (0.5 mg/kg) when administered in combination with ACY-738 or ACY-775 (Cit+ACY-738 vs veh: P<0.001; Cit+ACY-775 vs veh: P<0.01; Figure 4a). The potency of the effect of the combination with ACY-738 was comparable to that of a 40-fold higher dose of citalopram administered alone.
We previously reported a pro-resilient phenotype in the mouse CSD paradigm after serotonin-selective KO of HDAC6 (Espallergues et al, 2012). Here we tested whether chronic pharmacological inhibition of HDAC6 replicates this phenotype. Mice were pre-treated for 10 days with either vehicle, fluoxetine (20 mg/kg), ACY-738 (5 mg/kg), or ACY-775 (50 mg/kg), and were exposed to 10 days of CSD, with continued treatment (Figure 4c). On day 21, 24 h after last injection, mice were tested in the social interaction test. Compared with undefeated controls, vehicle-treated mice exposed to CSD showed a significant decrease in time in the interaction zone in the presence of a social target (Figure 4d; P<0.01) indicative of social avoidance. Treatment with fluoxetine, ACY-738 or ACY-775 prevented CSD-induced decrease in social interaction time. Defeated mice treated with ACY-738 showed significantly higher time in the interaction zone than those treated with vehicle (P<0.05). Similar trends were seen with ACY-775 (P=0.18) and fluoxetine (P=0.08), albeit these effects did not reach post-hoc significance. Similarly, treatment with ACY-738 led to a decrease in the amount of time spent in the corner zones (Figure 4e; P<0.05 vs vehicle). Significant effects were also observed when data were evaluated in terms of proportion of mice reaching criterion for resiliency. In mice treated with vehicle, we observed 27% of mice to be spontaneously resilient; treatment with fluoxetine increased the percentage to 60% (P<0.05); ACY-738 to 81% (P<0.01); and ACY-775 to 58% (P=0.12; Figure 4f). Drug treatments did not change total distance traveled (Figure 4g) or interaction time in undefeated control mice (data not shown). There was no significant effect of ACY-738 or ACY-775 on body weight gain during the 4 weeks of treatment.
This study establishes the in vitro selectivity, pharmacokinetics, and psychoactive properties of two novel brain-penetrant small molecule inhibitors of HDAC6, namely ACY-738 and ACY-775.
Interest in HDAC6 as a possible therapeutic target in the CNS (d'Ydewalle et al, 2012) has grown considerably following discovery of the key role of this enzyme in aggresome formation (Kawaguchi et al, 2003), pathological evidence for its inclusion in Lewy bodies in Parkinson’s disease as well as in glial cytoplasmic inclusions in frontotemporal lobar degeneration with TDP-43 inclusions (Ding et al, 2008), and evidence suggestive of a role in Alzheimer’s disease (Govindarajan et al, 2012; Sung et al, 2013; Xiong et al, 2013).
Tubastatin A and its newer analogs, which have an excellent HDAC6 selectivity, are currently the reference HDAC6 inhibitors. These probes have shown therapeutic promise in disease models of Charcot Marie Tooth disease (d'Ydewalle et al, 2011) and Alzheimer’s disease (Kim et al, 2012). To our knowledge, in vivo PK profiles of tubastatin A have not previously been reported. Our data suggest that after acute administration, at the dose tested here, tubastatin A has a limited brain biodistribution and does not lead to hyperacetylation of α-tubulin in CNS, the most abundant and best-characterized substrate of HDAC6. This relative lack of CNS activity upon acute systemic administration was in contrast to the potent effects of tubastatin A observed in neuronal culture and peripheral tissues. This suggests that behavioral and neuroprotective effects of tubastatin A previously reported in in vivo models of neurodegenerative disease (Benner et al, 2012) could be independent of α-tubulin acetylation or result from peripheral actions of the drug, such as the regulation of Foxp3+ regulatory T cells or other inflammatory mediators, which could indirectly affect neurodegenerative processes in brain (Vishwakarma et al, 2013; Beier et al, 2012).
For ACY-738 and ACY-775, we report an in vitro HDAC6 inhibitory potency in the low nanomolar range and an average selectivity of 100 × over class I isoforms. Despite their rapid elimination from plasma, biodistribution, and pharmacodynamic data indicate that upon systemic administration, ACY-738 and ACY-775 lead to sufficient brain exposure to serve as useful pharmacological probes of HDAC6 function in the brain. Indeed, our results show that both compounds induce dramatic changes in the acetylated fraction of α-tubulin persisting up to 4 h after acute systemic administration.
As predicted by our tissue culture results, the hyperacetylation of α-tubulin observed in vivo after administration of behaviorally active dose ACY-738 was not accompanied by detectable increases in H3K9 acetylation globally, as measured by western blot, or by enrichment of H3K9 binding at promoter regions of activity-induced neuronal genes (BDNF and cFOS), as measured by ChIP. In contrast, administration of a behaviorally active dose of the class I HDAC inhibitor sodium butyrate, which has been reported to increase global H3 acetylation (Gundersen and Blendy, 2009), resulted in elevations of H3 acetylation at these promoters. These results are consistent with the large body of literature including our own previous observations in HDAC6 KOs supporting a limited contribution of HDAC6 to chromatin remodeling. As behavioral effects of ACY-738 and ACY-775 are observed as soon as 15 min after treatment, these effects are unlikely to reflect the consequence of alterations in gene transcription and de novo protein synthesis. Nevertheless, based on a small number of studies showing that in certain cell types such as T cells HDAC6 is present in the nucleus and associates with chromatin and promoters, we cannot formally exclude the possibility that ACY inhibitors may also exert some of their activity through epigenetic regulation of CNS genes that were not directly examined here. Genome-wide analysis will be necessary to more definitively address this possibility.
It should be noted that despite its superior selectivity ACY-775 led to more variable results than ACY-738 in both the TST and social defeat paradigms. This is most likely attributable to 4.5-fold lower potency of ACY-775 and its more limited bioavailability possibly resulting in reduced target engagement. In addition, because of its poor solubility ACY-775 had to be given as a suspension at 50 mg/kg, another factor likely increasing individual variability in drug exposure.
Along with others, we have shown previously that serotonin neurons in the DRN are enriched in HDAC6 and provided evidence that genetic ablation of HDAC6, either globally or conditionally in serotonin neurons, induces antidepressant-like phenotypes (Espallergues et al, 2012; Fukada et al, 2012). Our present results confirm pharmacologically that inhibition of HDAC6 is a plausible antidepressant strategy.
Our results show that ACY-738 and ACY-775 display rapid psychoactive properties upon acute and chronic administration. We observed acute exploration-enhancing effects of these compounds in novel environments as well as anxiolytic-like effects in the open-field, marble-burying, and NIH tests, but a lack of effect in the EPM, where acutely administered antidepressants are typically inactive. We also detected acute antidepressant-like anti-immobility effects in the TST, as well as chronic pro-resilience effects in mice exposed to CSD. Acute behavioral effects in the open field and TST are fully attributable to the selective inhibition of HDAC6 in CNS, as these responses are occluded in mice with a conditional neural KO of HDAC6. Our observation of antidepressant-like effects of ACY-738 and ACY-775 in CSD model following chronic treatment with 5 mg/kg (ie, a dose inactive in acute models) suggest the implication of distinct neurobiological mechanisms or a possible potentiation of the therapeutic effects with repeated administration.
The ambulation-enhancing effects of ACY drugs observed here in the open field are similar to previously reported hyperlocomotor effect of serotonergic antidepressants (Brocco et al, 2002). As these effects are not observed when mice were tested in a home cage environment, they likely reflect an increased motivation in a novel environment or reduction in neophobia as opposed to a nonspecific enhancement of motor function. Two additional lines of evidence support this conclusion: first, significant effects of ACY-738 and ACY-775 were observed in the marble-burying test, where reduction in active response (ie, burying) is used as primary readout for drug activity. Second, antidepressant-like responses in CSD were observed in mice tested 24 h after the last administration, a time-point when the drug has been fully eliminated from the organism.
Our observations of antidepressant-like activity of HDAC6 inhibitors is in line with the results of previous studies showing that compounds such as SAHA, that have nM affinity for HDAC6, also produce antidepressant-like responses in mice (Covington et al, 2009). The observation that MS-275, a class I-selective inhibitor, produces antidepressant-like effects when infused in the brain suggests that both histone-dependent and -independent mechanisms likely contribute to the activity of these drugs (Covington et al, 2011; Golden et al, 2013).
Whether acetylation of α-tubulin, which is regulated by HDAC6 but not by class I HDACs, has a causal role in the activity of HDAC6 inhibitors remains an open question. It should be noted, however, that the class I selectivity of MS-275 does not necessarily imply an epigenetic selectivity in its mode of action, as large scale acetylation-proteomics studies have shown that MS-275 induces the hyperacetylation of numerous non-histone proteins (Choudhary et al, 2009), some of which are also hyperacetylated after ACY-738 administration or KO of HDAC6. The chaperone protein Hsp90, which has an important role in glucocorticoid receptor signaling, is one such common non-histone target of class I and class IIb HDACs that could plausibly mediate the common effects of class I and II HDAC inhibitors in stress resilience (Beier et al, 2012; de Zoeten et al, 2011; Espallergues et al, 2012; Nishioka et al, 2008).
Two major limitations of current antidepressants in clinical settings are the variability of response and delayed mode of action. This variability and delayed action are also observed in certain preclinical models like the CSD. Our results suggest that, when given prophylactically before and during exposure to CSD, HDAC6-selective inhibitors produce stress protective effects that appear more robust than those of the traditional SSRI fluoxetine. Furthermore, our observation that combination of a behaviorally inactive dose of ACY-738 with a subactive dose of citalopram produces behavioral effects equivalent to that of a 40-fold higher dose of citalopram administered alone, suggests a convergence of the downstream intracellular mechanisms engaged by SSRIs and HDAC6 inhibition. Our observations that a pan-neuronal HDAC6 KO potentiates SSRI activity suggests that HDAC6 blockade may act in part by lifting an inhibitory influence on serotonin-mediated signaling. This conclusion is further supported by a previous report showing that fluoxetine and global HDAC6 KO have synergistic antidepressant-like effects in mice (Fukada et al, 2012). Unexpectedly, we found that mice with Nestin-driven HDAC6 KO lack the baseline antidepressant and hyperlocomotor phenotype reported earlier in the global HDAC6 KO mice. These strain differences are likely to reflect cell-type and/or region-specific effects of HDAC6 and compensatory mechanisms dependent on the developmental period of the KO.
Taken together with the positive side-effect profiles of HDAC6 inhibitors (Raje et al, 2013), these results suggest that pharmacological HDAC6 inhibition may provide a useful approach to reduce rates of non-response and potentiate activity of currently available antidepressant treatments such as SSRIs.
Further studies will be necessary to better define the time courses and potential persistence of the antidepressant-like activity of HDAC6 inhibitors as well as their downstream mechanism of action. A likely mechanism may involve the Hsp90 chaperone complex and regulation of microtubule-dependent trafficking in axons and dendrites. Pharmacologic or genetic inhibition of HDAC6 has been shown to increase the trafficking of mitochondria (Chen et al, 2010) and dense core vesicles containing BDNF (Dompierre et al, 2007) in neurons. Along with others, we have shown that inhibition or genetic ablation of HDAC6 modulates nuclear trafficking of GR in response to glucocorticoid hormones and reduces stress signaling in DRN serotonin and prefrontal cortex pyramidal neurons, thereby promoting resilience (Espallergues et al, 2012; Lee et al, 2012). Tools developed and validated in this study will offer fundamentally novel opportunities to dissect the role of HDAC6 in cognitive and affective aspects of stress resilience.
FUNDING AND DISCLOSURE
OB is inventor on patent application #61713014 ‘Pyrimidine hydroxyamide compounds as protein deacetylase inhibitors and methods of use thereof ’. OB has received financial research support from Acetylon Pharmaceuticals. JHVD and MJ are employees of Acetylon Pharmaceuticals. RM is cofounder and member of the scientific advisory board of Acetylon Pharmaceuticals. RM has financial interests in SHAPE Pharmaceuticals and Acetylon Pharmaceuticals. He is also the inventor on IP licensed to these two entities. RM’s interests were reviewed and are managed by Massachusetts General Hospital and Partners HealthCare in accordance with their conflict of interest policies. JHVD and MJ and SJ are shareholders in Acetylon Pharmaceuticals. The remaining authors declare no conflict of interest.
We thank A Kozikowski for providing tubastatin A and I Lucki for his guidance and valuable input on the manuscript. This work was supported by grants MH087581 from the US National Institute of Mental Health and an award from the International Mental Health Research Organization (IMHRO) to OB and by training grant GM008076-29 from NIH.
About this article
JJ designed and performed the experiments, collected, analyzed and interpreted the data, and wrote the manuscript. JB performed the experiments. BGL designed and performed the experiments and interpreted data. JAB provided environment and support. MJ, JHVD, and RM contributed vital, novel reagents, and interpreted data. OB designed research, interpreted data, provided environment, and support and wrote the manuscript.
Supplementary Information accompanies the paper on the Neuropsychopharmacology website (http://www.nature.com/npp)