BACE inhibitor treatment of mice induces hyperactivity in a Seizure-related gene 6 family dependent manner without altering learning and memory

Abstract

BACE inhibitors, which decrease BACE1 (β-secretase 1) cleavage of the amyloid precursor protein, are a potential treatment for Alzheimer's disease. Clinical trials using BACE inhibitors have reported a lack of positive effect on patient symptoms and, in some cases, have led to increased adverse events, cognitive worsening and hippocampal atrophy. A potential drawback of this strategy is the effect of BACE inhibition on other BACE1 substrates such as Seizure-related gene 6 (Sez6) family proteins which are known to have a role in neuronal function. Mice were treated with an in-diet BACE inhibitor for 4-8 weeks to achieve a clinically-relevant level of amyloid-β40 reduction in the brain. Mice underwent behavioural testing and postmortem analysis of dendritic spine number and morphology with Golgi-Cox staining. Sez6 family triple knockout mice were tested alongside wild-type mice to identify whether any effects of the treatment were due to altered cleavage of Sez6 family proteins. Wild-type mice treated with BACE inhibitor displayed hyperactivity on the elevated open field, as indicated by greater distance travelled, but this effect was not observed in treated Sez6 triple knockout mice. BACE inhibitor treatment did not lead to significant changes in spatial or fear learning, reference memory, cognitive flexibility or anxiety in mice as assessed by the Morris water maze, context fear conditioning, or light-dark box tests. Chronic BACE inhibitor treatment reduced the density of mushroom-type spines in the somatosensory cortex, regardless of genotype, but did not affect steady-state dendritic spine density or morphology in the CA1 region of the hippocampus. Chronic BACE inhibition for 1-2 months in mice led to increased locomotor output but did not alter memory or cognitive flexibility. While the mechanism underlying the treatment-induced hyperactivity is unknown, the absence of this response in Sez6 triple knockout mice indicates that blocking ectodomain shedding of Sez6 family proteins is a contributing factor. In contrast, the decrease in mature spine density in cortical neurons was not attributable to lack of shed Sez6 family protein ectodomains. Therefore, other BACE1 substrates are implicated in this effect and, potentially, in the cognitive decline in longer-term chronically treated patients.

Figure 1

BACEi treatment decreases Aβ40 and Sez6 family protein ectodomain levels in the mouse brain and reduces cortical mushroom spine density. (A) The molecular structure of BACEi Compound H. (B) Aβ40 levels after 30 mg/kg/day BACEi treatment are reduced in WT and Sez6 TKO brains by 67% and 77% respectively as detected by Meso Scale Discovery assay. n = 15–16 per group; 2-way ANOVA: genotype p = 0.21, treatment p < 0.0001, interaction p = 0.21. (C) Sez6, Sez6L and Sez6L2 ectodomain protein fragments in brains from BACEi treated WT mice were decreased by 95%, 96% and 36% respectively compared to vehicle treated WTs as detected by Western blot. Representative Western blot (i) and quantitation of all samples (ii). n = 4–9/treatment; unpaired t-test: Sez6 p = 0.0012, Sez6L p = 0.0032, Sez6L2 p = 0.0018. Full-length blots are presented in Supplementary Figs. 2–4. (Di) Representative image of hippocampal dendrite segment used for analysis of dendritic spines. (Image shows a single Z plane however individual spines were examined in multiple planes as described in “Methods”). (Dii) BACEi treatment did not affect the density of dendritic spines on oblique apical secondary branches of hippocampal CA1 pyramidal neurons. The overall density and densities of individual spine classes were not altered by treatment in either WT or Sez6 TKO mice. n = 35 neurons (from 7 brains) per genotype/treatment; 2-way nested ANOVA of overall density: genotype p = 0.71, treatment p = 0.18, interaction p = 0.16. (Ei) Representative image of a cortical dendrite segment used for analysis of dendritic spines. (Eii) BACEi treatment did not affect the overall density of dendritic spines on basal secondary dendrites of layer V pyramidal neurons in the somatosensory cortex. However, BACE inhibition specifically reduced mushroom spine density in both WT and Sez6 TKO cortical neurons. WT vehicle: 0.41 ± 0.02 spines/µm, WT BACEi: 0.35 ± 0.02, TKO vehicle: 0.4 ± 0.02, TKO BACEi: 0.32 ± 0.02. n = 35 neurons (from 7 brains) per genotype/treatment; 2-way nested ANOVA of overall density: genotype p = 0.27, treatment p = 0.3, interaction p = 0.11; mushroom spine density: genotype p = 0.4, treatment p = 0.01, interaction p = 0.56. All graphs show mean ± SEM. **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

Figure 2

BACEi treatment does not alter spatial memory or cognitive flexibility in mice. (A) In the Morris water maze (MWM), WT and Sez6 TKO mice swam an equivalent distance to reach the hidden platform during acquisition training (from day 2) and reversal training (i) and Sez6 TKOs took a longer time than WT mice to find the hidden platform during reversal training (ii). 30 mg/kg/day BACEi treatment did not significantly affect pathlength or latency within genotypes (i–ii), or the ability of mice to successfully learn the hidden platform location in the acquisition and reversal probe trials (iii) although a tendency for BACEi treated WT mice to take longer to find the platform than vehicle treated WT mice on the first day of reversal training is noted (ii). n = 15–16 per group. (Ai) Log₁₀ of pathlength (mm) and (Aii) Log₁₀ of latency (seconds): 2-way RM ANOVA with acquisition and reversal analysed separately. (Ai) Acquisition: genotype p = 0.015, day p < 0.0001, treatment p = 0.73. (Ai) Reversal: genotype p = 0.891, day p < 0.0001, treatment p = 0.206. (Aii) Acquisition: genotype p = 0.091, day p < 0.0001, treatment p = 0.855. (Aii) Reversal: genotype p = 0.0004, day p < 0.0001, treatment p = 0.095. No significant interactions observed. Stars indicate a significant Bonferroni post-hoc at the level of genotype × day effect (treatment groups pooled). Aiii: 95% CI did not overlap with chance (7.5 s) for any group. (B) Sez6 TKO mice had enhanced fear learning at the 24 h context fear conditioning (CFC) retention test and delayed fear extinction compared to WT mice, but behaviour within each genotype was unaffected by 30 mg/kg/day BACEi treatment. n = 15–16 per group. 2-way RM ANOVA: genotype p < 0.0001, timepoint p < 0.0001, treatment p = 0.96, genotype × timepoint p < 0.0001, no other significant interactions. Stars indicate a significant Bonferroni post-hoc at the level of genotype × timepoint effect. (C) In the light/dark box test of anxiety, WT and Sez6 triple KO mice treated with 30 mg/kg/day BACEi spent an equivalent amount of time in the light zone (i) and made an equivalent number of entries into the light zone (ii). n = 15–16 per group. 2-way ANOVA. (Ci) genotype p = 0.62, treatment p = 0.16. (Cii) genotype p = 0.79, treatment p = 0.72. No significant interactions. All graphs show mean ± SEM except (Aiii) which shows mean ± 95% CI. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

Figure 3

BACEi treatment increases hyperactivity on the elevated open field in a Sez6 family protein dependent manner. (A) WT mice treated with 30 mg/kg/day BACEi moved a greater distance on the elevated open field (EOF) than vehicle treated mice, but this increased movement was not observed in BACEi treated Sez6 TKO mice. n = 16 per group. 2-way ANOVA: genotype p < 0.0001, treatment p = 0.036, interaction p = 0.21. Bonferroni post-hoc analysis: WT vehicle vs. WT BACEi p = 0.038, Sez6 TKO vehicle vs. Sez6 TKO BACEi p > 0.99. (B) WT mice treated with 15 mg/kg/day BACEi moved a greater distance on the EOF than vehicle treated mice, but this increased movement was not observed in BACEi treated Sez6 TKO mice. n = 12 per group. 2-way ANOVA: genotype p = 0.0001, treatment p = 0.067, interaction p = 0.045. Bonferroni post-hoc analysis: WT vehicle vs. WT BACEi p = 0.016, Sez6 TKO vehicle vs. Sez6 TKO BACEi p > 0.99. (C) A 20 mg/kg body weight intraperitoneal injection of cocaine immediately before testing on the EOF increased the distance travelled by WT and Sez6 TKO by 110% and 254% respectively. n = 9–11 per group. 2-way ANOVA: genotype p = 0.0017, treatment p < 0.0001, interaction p = 0.89. (D) Cocaine treatment before testing in locomotor cells increased the distance travelled by WT and Sez6 TKO by 187% and 281% respectively. n = 9–10 per group. 2-way ANOVA: genotype p = 0.0087, cocaine treatment p < 0.0001, interaction p = 0.75. Stars indicate a significant Bonferroni post-hoc analysis. All graphs show mean ± SEM. *p ≤ 0.05, ***p ≤ 0.001, ****p ≤ 0.0001.