Robust Translation of γ-Secretase Modulator Pharmacology across Preclinical Species and Human Subjects

Abstract

The amyloid-β peptide (Aβ)-in particular, the 42-amino acid form, Aβ1-42-is thought to play a key role in the pathogenesis of Alzheimer's disease (AD). Thus, several therapeutic modalities aiming to inhibit Aβ synthesis or increase the clearance of Aβ have entered clinical trials, including γ-secretase inhibitors, anti-Aβ antibodies, and amyloid-β precursor protein cleaving enzyme inhibitors. A unique class of small molecules, γ-secretase modulators (GSMs), selectively reduce Aβ1-42 production, and may also decrease Aβ1-40 while simultaneously increasing one or more shorter Aβ peptides, such as Aβ1-38 and Aβ1-37. GSMs are particularly attractive because they do not alter the total amount of Aβ peptides produced by γ-secretase activity; they spare the processing of other γ-secretase substrates, such as Notch; and they do not cause accumulation of the potentially toxic processing intermediate, β-C-terminal fragment. This report describes the translation of pharmacological activity across species for two novel GSMs, (S)-7-(4-fluorophenyl)-N2-(3-methoxy-4-(3-methyl-1H-1,2,4-triazol-1-yl)phenyl)-N4-methyl-6,7-dihydro-5H-cyclopenta[d]pyrimidine-2,4-diamine (BMS-932481) and (S,Z)-17-(4-chloro-2-fluorophenyl)-34-(3-methyl-1H-1,2,4-triazol-1-yl)-16,17-dihydro-15H-4-oxa-2,9-diaza-1(2,4)-cyclopenta[d]pyrimidina-3(1,3)-benzenacyclononaphan-6-ene (BMS-986133). These GSMs are highly potent in vitro, exhibit dose- and time-dependent activity in vivo, and have consistent levels of pharmacological effect across rats, dogs, monkeys, and human subjects. In rats, the two GSMs exhibit similar pharmacokinetics/pharmacodynamics between the brain and cerebrospinal fluid. In all species, GSM treatment decreased Aβ1-42 and Aβ1-40 levels while increasing Aβ1-38 and Aβ1-37 by a corresponding amount. Thus, the GSM mechanism and central activity translate across preclinical species and humans, thereby validating this therapeutic modality for potential utility in AD.

Fig. 1.

In vitro activity of BMS-932481. (A) H4-APPsw cell cultures were incubated overnight with BMS-932481 at a range of concentrations from 1 pM to 50 μM. Aβx-42 and Aβ1-x concentrations were determined simultaneously using the automated multiplex homogeneous time-resolved fluorescence assays. Error bars show standard error for seven independent assays. IC₅₀ values (or inhibition percentage) with standard deviations are as follows: Aβx-42 IC₅₀ = 5.5 ± 3.6 nM; Aβ1-x maximum inhibition = 30 ± 10% at 50 μM. (B) H4-APPsw cell cultures were incubated overnight with BMS-932481 at a range of concentrations from 10 pM to 10 μM. Aβ1-42 and Aβ1-40 were determined using automated homogeneous time-resolved fluorescence assays. Error bars show standard error for three independent assays in which both Aβ1-42 and Aβ1-40 were determined in parallel from the same cell cultures. IC₅₀ values with standard deviations are as follows: Aβ1-42 IC₅₀ = 6.6 ± 2.3 nM; Aβ1-40 IC₅₀ = 25 ± 7.9 nM. (C) H4-APPsw cell cultures were incubated overnight with BMS-932481 at a range of concentrations from 0.46 to 333 nM, 0.1% dimethylsulfoxide (DMSO) vehicle, or GSI BMS-299897 at a concentration of 1 μM. Concentrations were determined for Aβ1-42, Aβ1-40, Aβ1-38, and Aβ1-37 using the 4-plex Meso Scale Diagnostics assays. Error bars indicate standard error for two replicate wells. (D) An equimolar mix of synthetic peptides, including [¹⁴N]Aβ1-40, was evaluated by matrix-assisted laser desorption/ionization mass spectroscopy. (E and F) H4-APPsw cell cultures were incubated overnight with 0.1% dimethylsulfoxide (DMSO) (E) or BMS-932481 (F) at a concentration of 100 nM, then Aβ peptides were immunoprecipitated and evaluated by matrix-assisted laser desorption/ionization mass spectroscopy.

Fig. 2.

Altered levels of brain Aβ peptides in rats given single doses of BMS-932481. Rats were given intravenous doses of BMS-932481 at 1, 5, and 10 mg/kg, or vehicle alone. After dosing, groups (n = 4) of rats were euthanized at intervals between 10 minutes and 24 hours. Plasma, brain, and CSF samples were taken. Brain Aβ1-42 (A), brain Aβ1-40 (B), brain Aβ1-38 (C), brain Aβ1-37 (D), brain Aβ1-x (E), and the sum of brain Aβ1-42, Aβ1-40, Aβ1-38, and Aβ1-37 (F). (G) Stack chart showing amounts of Aβ1-42, Aβ1-40, Aβ1-38, and Aβ1-37 in brain and CSF of vehicle- and BMS-932481–dosed rats 7 hours after dosing. (H) Concentrations of BMS-932481 in blood plasma. Error bars indicate standard error. The concentrations of CSF Aβ1-42, Aβ1-40, Aβ1-38, Aβ1-37, and Aβ1-x in the same rats are shown in Supplemental Fig. 3. The significance of treatment effects was analyzed by analysis of variance (Supplemental Tables 1 and 2).

Fig. 3.

The effects of GSMs on Aβ levels in brain and CSF are correlated. ABEC was calculated for brain and CSF Aβ1-42, Aβ1-40, Aβ1-38, and Aβ1-37 at each dose of BMS-932481 and BMS-986133 in the rat study illustrated in Fig. 2 and Supplemental Figs. 3–5. (A) ABECs for CSF Aβ1-42 and CSF Aβ1-40 were plotted against the corresponding ABECs for brain Aβ1-42 and brain Aβ1-40. Linear regression showed a best fit of y = 0.99*x – 1.7, r² = 0.93, P < 0.0001. (B) ABECs for CSF Aβ1-38 and CSF Aβ1-37 were plotted against the corresponding ABECs for brain Aβ1-38 and brain Aβ1-37. Linear regression showed a best fit of y = 1.4*x + 14, r² = 0.98, P < 0.0001. (C) Scatter plot for brain Aβ1-42 and CSF Aβ1-42 from individual rats (total of 196 rats). Linear regression showed a best fit of y = 0.92*x + 7.9, r² = 0.34, P < 0.0001. (D) Scatter plot for brain Aβ1-40 and CSF Aβ1-40 from individual rats. Linear regression showed a best fit of y = 0.922*x + 14, r² = 0.61, P < 0.0001. (E) Scatter plot for brain Aβ1-42 and brain Aβ1-40 from individual rats. Linear regression showed a best fit of y = 0.82*x + 9.8, r² = 0.78, P < 0.0001. (F) Scatter plot for CSF Aβ1-42 and CSF Aβ1-40 from individual rats. Linear regression showed a best fit of y = 0.81*x + 19, r² = 0.85, P < 0.0001.

Fig. 4.

Altered levels of CSF Aβ peptides in dogs given single doses of BMS-932481. Dogs surgically fitted with a cannula in the lumbar spinal cord were given oral doses of BMS-932481 at 2, 5, and 30 mg/kg, or vehicle alone in a cross-over study design (total of four dogs). Plasma and CSF samples were taken predose and at intervals after dosing up to 72 hours. CSF Aβ1-42 (A), CSF Aβ1-40 (B), CSF Aβ1-38 (C), CSF Aβ1-37 (D), and the sum of CSF Aβ1-42, Aβ1-40, Aβ1-38, and Aβ1-37 (E). (F) Scatter plot for CSF Aβ1-42 and CSF Aβ1-40 from all individual samples (total of 189 samples). Linear regression showed a best fit of y = 0.81*x + 20, r² = 0.77, P < 0.0001. Error bars indicate standard error. Concentrations of BMS-932481 in blood plasma from the dogs are illustrated in Supplemental Fig. 6. The significance of treatment effects was analyzed by analysis of variance (Supplemental Table 3).

Fig. 5.

Altered levels of CSF Aβ peptides in monkeys given single doses of BMS-986133. Monkeys surgically fitted with a cannula in the lumbar spinal cord were given intravenous doses of BMS-932481 at 5 and 15 mg/kg, or vehicle alone, in a crossover study design (total of four monkeys). Plasma and CSF samples were taken predose and at intervals after dosing up to 72 hours. CSF Aβ1-42 (A), CSF Aβ1-40 (B), CSF Aβ1-38 (C), CSF Aβ1-37 (D), CSF Aβ1-x (E), and the sum of Aβ1-42, Aβ1-40, Aβ1-38, and Aβ1-37 (F). (G) Concentrations of BMS-932481 in blood plasma. (H) Scatter plot for CSF Aβ1-42 and CSF Aβ1-40 from all individual samples (total of 108 samples). Linear regression showed a best fit of y = 0.86*x + 15, r² = 0.97, P < 0.0001. Error bars indicate standard error. The significance of treatment effects was analyzed by analysis of variance (Supplemental Table 4).

Fig. 6.

Altered levels of CSF Aβ peptides in human subjects given a single oral dose of BMS-932481. Healthy human subjects were given a single 900-mg oral dose of BMS-932481 (n = 10) or placebo (n = 5), and CSF samples were taken through an implanted lumbar catheter at intervals up to 24 hours. CSF Aβ1-42 (A), CSF Aβ1-40 (B), CSF Aβ1-38 (C), CSF Aβ1-37 (D), CSF Aβ1-x (E), and the sum of Aβ1-42, Aβ1-40, Aβ1-38, and Aβ1-37 (F). (G) Aβ1-42 and Aβ1-x were calculated as a percentage relative to predose levels, then each Aβ1-42 value was divided by the corresponding Aβ1-x value. (H) Scatter plot for CSF Aβ1-42 and CSF Aβ1-40 from all individual samples (total of 108 samples). Linear regression showed a best fit of y = 0.81*x + 17, r² = 0.96, P < 0.0001. Error bars indicate standard error. The significance of treatment effects was analyzed by analysis of variance (Supplemental Table 5).

Fig. 7.

Alignment of BMS-932481 PK/PD across species. (A) Scatter plot of CSF Aβ1-42 trough versus plasma AUC: CSF Aβ1-42 trough (minimum level of Aβ1-42 after dosing) and plasma AUC were determined for each individual human subject, and for each dose in individual dogs and monkeys. For rats, Aβ1-42 trough for brain, Aβ1-42 trough for CSF, and plasma AUC were calculated using group means. Rat values were derived from the experiment illustrated in Fig. 2, Supplemental Figs. 3–5, and two additional time-course studies with BMS-932481 (not shown). Nonlinear fit of the entire data set indicates 50% inhibition at AUC = 11 μM⋅h. (B) Scatter plot of CSF Aβ1-42 trough versus plasma Cmax. Nonlinear fit indicates 50% inhibition at C_max = 1.5 μM.