Presenilin-1 knockin mice reveal loss-of-function mechanism for familial Alzheimer's disease

Abstract

Presenilins play essential roles in memory formation, synaptic function, and neuronal survival. Mutations in the Presenilin-1 (PSEN1) gene are the major cause of familial Alzheimer's disease (FAD). How PSEN1 mutations cause FAD is unclear, and pathogenic mechanisms based on gain or loss of function have been proposed. Here, we generated Psen1 knockin (KI) mice carrying the FAD mutation L435F or C410Y. Remarkably, KI mice homozygous for either mutation recapitulate the phenotypes of Psen1(-/-) mice. Neither mutation altered Psen1 mRNA expression, but both abolished γ-secretase activity. Heterozygosity for the KI mutation decreased production of Aβ40 and Aβ42, increased the Aβ42/Aβ40 ratio, and exacerbated Aβ deposition. Furthermore, the L435F mutation impairs hippocampal synaptic plasticity and memory and causes age-dependent neurodegeneration in the aging cerebral cortex. Collectively, our findings reveal that FAD mutations can cause complete loss of Presenilin-1 function in vivo, suggesting that clinical PSEN mutations produce FAD through a loss-of-function mechanism.

Figure 1. L435F and C410Y KI/KI mice exhibit phenotypes indistinguishable from Psen1^−/− mice

(A) Targeting strategy for the generation of L435F and C410Y KI mice. Boxes represent Psen1 exons 8–12, and the locations of the 5′ and 3′ external probes are shown. The L435F mutation is located in exon 12 and C410Y mutation is located in exon 11. Ba and Ap denote BamHI and ApaI restriction sites, respectively. (B) Newborn L435F and C410Y KI/KI pups resemble Psen1^−/− pups with shortened tail and body axis. (C) Northern analysis of brain total RNA at P0 shows normal Psen1 mRNA size and level in L435F and C410Y KI/KI and KI/+ mice (n≥5). (D) Western analysis of brain lysates at P0 shows residual amounts of PS1 NTFs and CTFs in L435F and C410Y KI/KI brains and drastic accumulation of PS1 holoprotein in L435F and C410Y KI/KI and KI/+ brains in a KI gene dosage-dependent manner, whereas these PS1 species are undetectable in Psen1^−/− brains (n≥4). Data are represented as mean ± SEM. ***p < 0.001. See also Figure S1 and Table S1.

Figure 2. Impaired neurogenesis and Notch signaling in L435F and C410Y KI/KI mice

(A) In vitro γ-secretase assay using CHAPSO-solubilized E18.5 brain fractions and recombinant Notch substrate N102-FmH followed by Western analysis reveals that NICD production is reduced in L435F KI/+ (n=3) and C410Y KI/+ (n=4) mice, and abolished in L435F KI/KI (n=3) and C410Y KI/KI (n=3) mice, compared to Psen1^+/+ controls (n=6). SNAP25 is used as internal loading control for membrane fractions. (B) Northern analysis of Hes5 expression. Levels of Hes5 mRNA are markedly reduced in L435F KI/KI (n=9), C410Y KI/KI (n=3), Psen1^−/− (n=3) brains relative to Psen1^+/+ controls (n=7). (C) Hematoxylin/eosin staining of comparable brain sections from L435F KI/KI, C410Y KI/KI and littermate controls at E16.5 shows that the ventricular zone (VZ) and the developing cortical plate (CP) are thinner in L435F KI/KI and C410Y KI/KI mice. The bottom panels show higher power views of the boxed areas in the developing CP. Dashed lines show the boundaries of the CP. Scale bar: 0.25mm. (D) The number of proliferating neural progenitor cells in the VZ labeled by Ki67 immunoreactivity is reduced in L435F KI/KI and C410Y KI/KI brains at E16.5 (n=4 for each genotype). Scale bar: 0.25mm. Data are represented as mean ± SEM. ***p < 0.001.

Figure 3. Abolished Aβ production in L435F KI/KI mice and reduced Aβ40 and Aβ42 production in L435F KI/+ mice

(A) APP CTFs accumulate (>10-fold) in L435F KI/KI and Psen1^−/− brains relative to littermate controls at E18.5 (n≥6). (B) ELISA measurements of Aβ40 and Aβ42 following in vitro γ-secretase assay. In vitro γ-secretase assay reveals that de novo generation of Aβ40 and Aβ42 in L435F KI/+ brains at E18.5 (n=4) is reduced by 46% and 43%, respectively, compared to Psen1^+/+ brains (n=5), and they are undetectable in L435F KI/KI brains (n=3). (C) ELISA measurements of endogenous Aβ40 and Aβ42 from brain homogenates at E18.5. Levels of stead-state endogenous Aβ40 and Aβ42 are similar between L435F KI/+ and Psen1^+/+ brains, and they are undetectable in L435F KI/KI brains (n=4 for each genotype). (D) ELISA measurements of Aβ40 and Aβ42 following in vitro γ-secretase assay in 3-month old mice. De novo generation of Aβ40 and Aβ42 in L435F KI/+ cortical homogenates is reduced by 47% and 43%, respectively, compared to Psen1^+/+ cortices (n=6 for each genotype). (E) ELISA measurements of stead-state endogenous Aβ40 and Aβ42 in both insoluble and soluble fractions from cortical samples at 3 months of age. In insoluble fractions, levels of endogenous Aβ40 and Aβ42 are reduced in L435F KI/+ cortices (n=10) relative to controls (n=7), and the Aβ42/Aβ40 ratio is increased. In soluble fractions, levels of endogenous Aβ40 and Aβ42 are not significantly different between Psen1^L435F/+ and Psen1^+/+ cortices, but the Aβ42/Aβ40 ratio is significantly increased in Psen1^L435F/+ cortices. Data are represented as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S2 and Table S2.

Figure 4. Reduced human Aβ accumulation but accelerated amyloid deposition in L435F KI/+ mice expressing a human mutant APP transgene

(A) ELISA measurements of steady-state human Aβ40 and Aβ42 in insoluble and soluble fractions from Psen1^L435F/+; APP^WT (n=4) and controls (n=4) at 9 months of age. In Psen1^L435F/+; APP^WT mice, human Aβ40 is reduced by ~26% in insoluble fractions and ~62% in soluble fractions, whereas human Aβ42 is reduced by ~19% in insoluble fractions but not significant changed in soluble fractions. Note that levels of Aβ40 and Aβ42 are much higher in insoluble fractions than in soluble fractions (>20-fold for Aβ40, >10-fold for Aβ42). The Aβ42/Aβ40 ratio is significantly enhanced in soluble and insoluble fractions. (B) ELISA measurements of steady-state human Aβ40 and Aβ42 in insoluble and soluble fractions of cortical homogenates from Psen1^+/+; APP^MT (n=4) and Psen1^L435F/+; APP^MT (n=4) at 2 months of age. Levels of human Aβ40 are significantly reduced in insoluble fractions but not in soluble fractions, whereas levels of human Aβ42 are not significantly altered. Note that levels of Aβ40 are drastically higher in insoluble fractions than in soluble fractions (>100-fold). The Aβ42/Aβ40 ratio is significantly increased in insoluble fractions. (C) Aβ immunostaining with an Aβ antibody (R1282) reveals significantly increased plaque deposition in the cerebral cortex of Psen1^L435F/+; APP^MT mice compared to Psen1^+/+; APP^MT mice at 9 months of age. Scale bar: 0.25 mm. Data are represented as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S3 and Table S2.

Figure 5. Memory impairment caused by the L435F mutation

(A) Psen1^L435F/+; Psen2^−/− (KI/+; −/−) mice (n=14) showed significantly higher escape latencies than Psen1^+/+; Psen2^−/− (+/+; −/−) littermates (n=15) during the 14-day training period (F_{1, 27} = 8.06; p<0.01) in the Morris water maze test. (B) Psen1^L435F/+; Psen2^−/− mice show significantly reduced target quadrant occupancy in the day 7 probe test, but both genotypic groups show similar target quadrant occupancy in the day 13 probe test. Psen1^L435F/+; Psen2^−/− mice display significantly reduced quadrant occupancy in the partial-cue probe test on day 14. T, target quadrant; AL, adjacent left quadrant; AR, adjacent right quadrant; OP, opposite quadrant. (C) Naïve Psen1^L435F/+; Psen2^−/− mice (n=10) and littermate controls (n=9) were subjected to 8 days of training to locate the three baited arms (indicated by red *) following fasting in the radial arm maze test. (D) Psen1^L435F/+; Psen2^−/− and control mice require similar amounts of time to complete the food search (F_{1, 17} = 1.18, p > 0.05) during the 8-day radial arm maze training. (E) Psen1^L435F/+; Psen2^−/− mice display significantly more reference errors (F_{1, 17} = 4.59, p < 0.05) during the 8-day training period. (F) Psen1^L435F/+; Psen2^−/− mice show significantly higher percentages of 45 degree turns (F_{1, 17} = 13.63, p < 0.001) during the 8-day training period. Data are represented as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S4.

Figure 6. Synaptic dysfunction caused by the L435F mutation

(A) Normal basal synaptic transmission in Psen1^L435F/+; Psen2^−/− mice. The synaptic input-output relationship was obtained by plotting the fiber volley amplitude against the initial slope of the evoked fEPSP. Each point represents data averaged across all slices for a narrow bin of FV amplitude. The lines represent the best linear regression fit. (B) Reduced paired-pulse facilitation (PPF) in Psen1^L435F/+; Psen2^−/− mice. The graph depicts the paired-pulse response ratio obtained at different inter-stimulus intervals (in ms). Scale bar: 1 mV, 50 ms. (C) Synaptic facilitation elicited by stimulus trains of the indicated frequencies in Psen1^L435F/+; Psen2^−/− and littermate control mice. Frequency facilitation is significantly decreased in the 5–20 Hz range in Psen1^L435F/+; Psen2^−/− mice. Scale bar: 1 Hz: 1 mV, 500 ms; 20 Hz: 1 mV, 50 ms. FP, field potential. (D) Pairing-induced LTP at the Schaffer collateral-CA1 synapses in slices from Psen1^L435F/+; Psen2^−/− mice and littermate controls. Representative traces before (thin) and after (thick) LTP induction are shown. Insets show the average of 15 EPSCs recorded before and 45–50 min after the induction. Scale bar: 20 ms, 50 pA. (E) LTP induced at the recurrent commissural/associational (C/A)-CA3 synapses in slices from Psen1^L435F/+; Psen2^−/− mice and littermate controls. LTP was induced with 3 trains of high frequency stimulation (HFS). Representative traces before (thin) and after (thick) LTP induction are shown. Superimposed traces are averages of four consecutive responses 1 min before and 50 min after HFS. The lack of response to DCG IV, which selectively inhibits mossy fiber LTP, confirms that the responses are obtained from (C/A)-CA3 synapses. Scale bar: 10 ms, 0.5 mV. (F) Increased synaptic depression at (C/A)-CA3 synapses during LTP-inducing stimulation in Psen1^L435F/+; Psen2^−/− mice. Insets show representative fEPSP traces obtained during single stimulation train. Scale bar, 10 ms, 2 mV. The values in parentheses indicate the number of neurons or slices (left) and the number of mice (right) used in each experiment. Data are represented as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S5.

Figure 7. The L435F mutation causes age-dependent neurodegeneration

(A) Comparison of brain morphology between Psen1^F/L435F; Psen2^−/−; Cre mice and littermate controls at 2, 12 and 18 months of age. Left: Images of Nissl stained comparable sagittal brain sections. Right: Cortical volume and neuron number obtained by stereological quantification following Nissl staining and NeuN immnostaining, respectively. Cortical volume and neuron number are normal at 2 months of age but significantly reduced at 12 (~22%) and 18 (~32%) months of age in Psen1^F/L435F; Psen2^−/−; Cre mice relative to controls (n≥3 for each genotype at each age). Scale bar: 0.1 mm. (B) More TUNEL-positive cells are observed in the cerebral cortex of Psen1^F/L435F; Psen2^−/−; Cre mice (n=3), compared to control mice (n=4) at 18 months of age. Scale bar: 0.1 mm. (C) More active caspase3-positive cells are detected in the cerebral cortex of Psen1^F/L435F; Psen2^−/−; Cre mice (n=3), relative to control mice (n=4) at 18 months of age. Scale bar: 0.1 mm. (D) Increased GFAP immunoreactivity in the cortex of Psen1^F/L435F; Psen2^−/−; Cre mice at 12 months of age indicates astrogliosis. Bottom images show higher magnification views of the boxed area in the cortex. Western analysis also shows increased levels of GFAP in the cortex of Psen1^F/L435F; Psen2^−/−; Cre mice (KI) compared to littermate control Psen1^F/F; Psen2^−/− and Psen1^F/+; Psen2^−/−; Cre mice (n=8 for each genotype). Data are represented as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001. Scale bar: 0.1mm. See also Figure S6.

Figure 8. A schematic model for FAD

Pathogenic PSEN1 mutations cause loss of PS1 function and γ-secretase activity, leading to AD-relevant functional and neuropathological changes, including memory impairment, synaptic dysfunction, age-dependent neurodegeneration, gliosis and dementia. In parallel, the loss of PS function produced by pathogenic PSEN1 mutations also reduces Aβ production and increases the Aβ42/Aβ40 ratio due to the greater reduction of Aβ40 production than Aβ42, thereby promoting amyloid plaque deposition. Thus, PSEN mutations produce the full spectrum of AD phenotypes through a loss-of-function mechanism.