At the end of the study, mice were sacrificed at putative brain Tmax (1 h post last dose) based on the previous PK observations. Cortical tissue from the mice was assessed using specific antibodies for increases in acetylation of specific lysine residues on Histones H3 (H3K4, H3K9, H3K14) and H4 (H4K5) as well as global acetylation of H3 (Ac-H3) and H4 (Ac-H4). The acetylation level was normalized to total H3 and H4 expression level. The samples from the rest of the brain were analyzed for terminal 4b levels at the time of sacrifice. As a positive control, a separate cohort of mice was injected sc with 200 mg/kg SAHA according to previous procedures [22] and sacrificed at the same time for evaluation of histone acetylation. As illustrated in Fig. 5, dosing of 4b at 150 mg/kg po bid for five consecutive days failed to enhance histone acetylation. In contrast, a robust enhancement of acetylation of H3K9, H4K5, and total acetylated levels of H3 and H4 were observed with SAHA, in line with previous observations [22]. In agreement with the lack of an observable central pharmacodynamic response to the administration of 4b, concentrations of 4b in the “rest of the brains” collected (n = 12 of 14) were below the limit of quantitation (LLQ was 5 nM), in this study in all but two animals examined. These animals demonstrated low but detectable brain concentrations at 20 and 23 nM, respectively, which may be attributed to residual blood in the brain of these non-perfused animals. These exposure data are in good agreement with the previous PK analysis showing negligible brain penetration of the compound with 50 mg/kg po.
Discussion
Due to the published in vitro and in vivo results with the pimelic diphenylamide HDAC inhibitors in cell and mouse models of FRDA and HD, we evaluated the HDAC isoform selectivity, cellular activity, in vitro and in vivo ADME properties of the preclinical prototype compound HDACi 4b to validate and extend previous findings and assess its therapeutic potential for HD. Our data on the in vitro selectivity and binding mode of this compound largely agree with previous reports for the closelyrelated analogue 106, which demonstrates a unique slow-on/slowoff binding mode of these HDAC Class I inhibitors relative to hydroxamic acid-based HDAC inhibitors. The association of 106 to HDAC3 association was previously reported to proceed considerably more slowly than the association to HDAC1. Dissociation rates of the complex also differed; for the 106:HDAC3 complex, the half-life (at room temperature) was ,6 h, whereas it was ,1.5 h for the 106:HDAC1 complex [36]. In that study, a progression method was used to assess the Ki values of 106 binding to HDACs, resulting in a kinetic profile for 106 of HDAC3 (Ki 14 nM) . HDAC2 (Ki 102 nM) . HDAC1 (Ki 148 nM), quite different from the selectivity profile of HDAC1 (IC50 = 150 nM) . HDAC3 (IC50 = 370 nM) . HDAC2 (IC50 = 760 nM), measured using conventional assays with compound-enzyme incubation times of 1 to 3 h [36]. Bressi et al have since proposed a model (following crystal structure determination of a ortho-N-acyl-phenylene diamine inhibitor bound to HDAC2) in which disruption of an intramolecular hydrogen bond of the NH2 group to the carbonyl oxygen is required for this tight binding and could be responsible for the slow-on/slow-off kinetics [37]. A publication by Xu et al further defined 106 activity as being highly preferential for HDAC3 inhibition over HDAC1 and HDAC2.
group attached through a flexible ethylene glycol linker to 106 plus an alkyne group for subsequent attachment of an azide-linked reporter dye for affinity capture [38]. 1-BP retained HDAC inhibitory activity against recombinant HDACs 1, 2, and 3 equivalent to 106. 1-BP was subsequently used for HDAC isoform target identification when incubated with recombinant HDACs followed by irradiation to effect photo cross-linking, fluorescent dye attachment by click chemistry and gel electrophoresis. The 1-BP: HDAC3 interaction was by far the strongest association seen, with much lower association of 1-BP: HDAC1 being the only other HDAC interaction noted, and only when higher enzyme concentrations were used. This result [38] appeared at odds with the earlier publication [36] reporting good HDAC1 and HDAC2 inhibition following prolonged incubation of enzyme with benzamide 106. The authors speculated that the increased stability of the 106:HDAC3 complex accounted for the difference in cross-linking activity of 1BP for these enzymes, and concluded that HDAC3 was the preferred cellular target of the pimelic diphenylamide inhibitor 106 used in the in vivo FRDA mouse models, which is very closely related in structure to 4b used in the R6/2 HD mouse model [38]. They also speculated that the efficacy of the pimelic diphenylamide inhibitors versus the lack of efficacy of the hydroxymate inhibitors to increase FXN expression in published reports was due to the absolute requirement of this stable HDAC3: inhibitor complex. In our study, our `functional’ deacetylase inhibition data shows a modest selectivity for HDAC3 over HDAC1 and HDAC2, which in our opinion is not a sufficient pharmacological basis to identify HDAC3 as the exclusive target for 4b actions in vivo. In addition to reconfirming the biochemical profile of 4b as exemplifying the novel mode of action (slow-on/slow-off) of pimelic diphenylamide HDAC inhibitors for Class I HDACs, our data provides insight into the cellular inhibitory profile of endogenous Class I HDAC inhibition and helps define the concentrations required in plasma or brain to effectively inhibit the target in a native cell environment. In our study, a maximal cellular IC50 for `Class I’ HDAC inhibition of 1.8 mM after 24 h incubation was achieved and used to benchmark in vivo central target engagement. Thus, an in vivo proof of concept study linking functional central inhibition of Class I HDACs to efficacy in a disease state should ideally provide confirmation of at least some correlation between significant target engagement and phenotypic outcome, to support pursuit of this approach in a clinically afflicted population. We further characterized the in vitro ADME profile of 4b to explore any metabolic liabilities that would inform subsequent in vivo efficacy testing and dosing schedule. Our findings indicate that 4b is very unstable in plasma (T1/2 1.9 hr) and in liver microsomes (T1/2 with NADPH as cofactor ,20 min) resulting in a very high predicted in vivo plasma clearance of 2.6 L/h/kg (approximately 87% liver plasma flow). The main hydrolysis product in plasma is M1, which accounts for 78% of the metabolism of parent, followed by M2, which accounts for 10% of the entire metabolism. Our biochemical and cellular profiling of these two metabolites confirmed that neither inhibits HDACs in a cellular setting (up to 50 mM), confirming that these metabolic products of 4b would not inhibit HDACs in vivo. In summary, based on our in vitro ADME data, we predict that 4b will be rapidly metabolised by plasma and hepatic amidases and other hepatic enzymes in vivo, significantly limiting systemic and CNS availability. Additionally, we show that 4b is a substrate for Pgp in MDCK cells overexpressing MDR1 (efflux ratio (ER) of 4.9), in agreement with the efflux determined in Caco2 monolayers (ER 3.5). This in vitro data predicts a limited
Figure 5. 4b treatment does not affect histone acetylation in mouse brain. (A) Representative immunoblot showing histone acetylation in mouse brain in response to 4b treatment. Mice treated with SAHA were used as a positive control (B). Acetylation at specific lysine residues on histone 3 (H3K4, H3K9, H3K14) and Histone 4 (H4K5) as well as global acetylation of H3 (Ac-H3) and H4 (Ac-H4) were studied using specific antibodies.