|Home | About | Journals | Submit | Contact Us | Français|
Mutations in the presenilin-1 (PSEN1) gene are associated with familial Alzheimer's disease and frontotemporal dementia (FTD). Interestingly, neuropathological analysis of a Belgian FTD family carrying a PSEN1 c.548G>T mutation confirmed neurodegeneration in the absence of amyloid plaques. To investigate the impact of the c.548G>T mutation on presenilin-1 (PS1) function in vivo, we introduced this mutation into the genomic Psen1 locus. The resulting c.548G>T knockin (KI) mice are viable but express markedly lower levels of Psen1 mRNA and protein in the brain. This reduction is due to production of aberrantly spliced transcripts lacking either exon 6 or exons 6 and 7 and their subsequent degradation via nonsense-mediated decay (NMD); inhibition of NMD by cycloheximide treatment stabilized these transcripts and restored the level of Psen1 mRNA in KI/KI brains. Interestingly, the reduction of Psen1 mRNA expression and the degradation of aberrant Psen1 splice products occur exclusively in the brain but not in other tissues. Consistent with decreased Psen1 expression, γ-secretase activity was strongly reduced in the cerebral cortex of KI mice, as measured by de novo γ-secretase-mediated cleavage of APP and Notch. Moreover, PS1 expressed from Psen1 cDNA carrying the c.548G>T mutation displayed normal γ-secretase activity in cultured cells, indicating that the corresponding p.183G>V amino acid substitution does not affect γ-secretase activity. Finally, Psen1 c.548G>T KI/KI; Psen2−/− mice exhibited mild spatial memory deficits in the Morris water maze task. Together, our findings demonstrate that the c.548G>T mutation results in a brain-specific loss of presenilin function due to decreased Psen1 mRNA expression.
Frontotemporal dementia (FTD) is the second most common form of dementia following Alzheimer's disease (AD) (Sjogren and Andersen, 2006). FTD is a clinically diverse syndrome characterized by profound behavioral changes and degeneration of the frontal and anterior temporal cortex (Neary et al., 2005). Due to the clinical and neuropathological heterogeneity, FTD comprises a number of related disorders with overlapping but distinct features, including Pick's disease and frontotemporal lobar degeneration. For example, tau pathology is present in some FTD cases whereas others lack tau deposition. A large proportion (~20–50%) of FTD cases have a familial component, and mutations in microtubule-associated protein tau and progranulin genes are the most frequent genetic causes (Galimberti and Scarpini, 2010). Interestingly, mutations in the presenilin-1 (PSEN1) and presenilin-2 (PSEN2) genes, which are the major cause of familial AD, have also been implicated in FTD (Mendez and McMurtray, 2006). Since the first identification of the PSEN1 c.338T>C mutation in cases of familial dementia with prominent clinical frontotemporal features (Raux et al., 2000), more than 10 mutations in the PSEN genes have been associated with clinical diagnoses of FTD, in some cases accompanied by frontotemporal atrophy and frontotemporal hypoperfusion on neuroimaging studies (Rippon et al., 2003; Dermaut et al., 2004; Halliday et al., 2005; Zekanowski et al., 2006; Bernardi et al., 2009; de Bot et al., 2009; Marcon et al., 2009; Gallo et al., 2010; Borroni et al., 2011). Moreover, some pathogenic PSEN1 mutations can cause neuropathological changes consistent with co-existing Pick's disease and AD (Ikeda et al., 1996; Halliday et al., 2005). Interestingly, several of these mutations reside at the exon/intron boundaries, and therefore may affect PSEN1 splicing in addition to causing missense substitutions (Raux et al., 2000; Dermaut et al., 2004; Borroni et al., 2011). Thus, mutations in PSEN may result in an overlapping clinical and neuropathological manisfestations of AD and FTD, and functional changes of Presenilin (PS) may underlie common pathogenesis of both dementias.
To investigate how PSEN mutations may be associated with FTD, we chose the PSEN1 c.548G>T mutation, which was originally identified in familial FTD patients with neuropathological confirmation of Pick's-type tauopathy in the absence of amyloid deposition (Dermaut et al., 2004). Since the c.548G>T mutation resides at the last nucleotide of exon 6, we generated knockin (KI) mice in which the c.548G>T mutation was introduced into the genomic Psen1 locus. The c.548G>T KI/KI mice are viable, but Psen1 mRNAs are significantly decreased selectively in the brain due to aberrant exon skipping and subsequent degradation of aberrantly spliced transcripts by nonsense mediated mRNA decay. Accordingly, γ-secretase activity is reduced in the KI brain, as measured by γ-secretase-mediated cleavage of two physiological substrates, Notch and APP. However, PS1 expressed from full-length Psen1 c.548G>T mRNA displayed normal γ-secretase activity when tested in cultured cells. Furthermore, Psen1 KI/KI; Psen2−/− mice exhibited a significant deficit in spatial reference memory. Together, these findings demonstrate that the c.548G>T mutation causes brain-specific reduction of Psen1 mRNA expression and PS function in the maintenance of γ-secretase activity and memory.
For the Psen1 c.548G>T KI construct, a 2.49-kb left-arm fragment and a 3.15-kb right-arm fragment surrounding exon 6 were amplified by PCR using BAC DNA harbouring the mouse Psen1 gene (clone RP23-330F11, Children's Hospital Oakland Research Institute) as a template. The primer sequences are 5'-TACCGCGGAATGGGATGTGTGTGTTGGGATGC-3' and 5'-TGGCGGCCGCATGTGAGAATCCTGGGTGCAGTC-3' for the left-arm (the underlined sequences are for SacII and NotI, respectively), and 5'-GCGTCGACAAGTATGTGCTGATCCCCAAAGC-3' and 5'-GCAAGCTTAAGTGCTGGGATTACAGGAGGAC-3' for the right-arm (the underlined sequences are for SalI and HindIII recognition sites, respectively). The c.548G>T pathogenic mutation (encoding a missense mutation p.183G>V) and c.546A>G humanized nucleotide change (a silent mutation for Leu at the amino acid residue 182) were introduced by site-directed mutagenesis into exon 6 in the right-arm fragment (Fig. 1A, only c.548G>T mutation will be described hereafter). The left- and right-arm fragments were then subcloned into the SacII-NotI site and the SalI-HindIII site of the PGKneolox2DTA plasmid (gift of P. Soriano), respectively, to produce the KI targeting vector (Fig. 1B), which was then confirmed by extensive restriction digestions and sequencing. The linearized targeting vector by XhoI was electroporated into MKV6.5 ES cells (gift of R. Jaenisch), and selected with 150 μg/ml G418 (Invitrogen) for 4–5 days. Genomic DNAs were purified from total 360 independent ES clones, and subject to Southern analysis using the 5' probe. For Southern analysis, genomic DNAs were digested with BamHI. Among 52 clones that were tested positive of proper homologous recombination in the 5' arm, 3 ES clones (3-9-H, 4-1-F, 7-11-A) were expanded and subjected to further Southern analysis using the 5', 3', and neomycin probes (Fig. 1B). We confirmed that homologous recombination correctly occurred in both the 5' and 3' homologous regions in these 3 clones. To generate Psen1 c.548G>T KI mice, the targeted ES cells (clones 4-1-F and 7-11-A) were injected into C57BL/6 blastocysts, and the resulting male chimeric mice were mated with C57BL6/J-129 F1 mice. Mice transmitting the targeted allele were further crossed to αCaMKII-Cre male transgenic mice (Yu et al., 2001) to excise the floxed PGK-neo cassette, as the αCaMKII promoter allows the Cre transgene to express weakly in male germ cells (Ignotz and Suarez, 2005). The resulting progeny (KI heterozygous mice) were intercrossed to obtain homozygous KI mice for further characterization. Correct homologous and Cre-mediated site-specific recombination events were further confirmed by Southern analysis with tail genomic DNAs using the external 5', 3' and the neo probes, and by sequencing to detect the presence of the c.548G>T mutation in genomic DNAs. Since offsprings from both targeted ES clones 4-1-F and 7-11-A were indistinguishable, we utilized the KI mice derived from the ES clone 4-1-F for further characterisation. In subsequent generations, mice were genotyped routinely by PCR and the c.548G>T KI allele was detected by size shift due to the remaining loxP and construct sequence between exons 5 and 6 using the following primers, 5'-TGGTGAGAGCTCAGCAGGTAAG-3' and 5'-TGCTTTCTAGTTGTCCTTCGTCG-3'. The 410 bp band represents the wild-type allele whereas the 529 bp band represents the KI allele. Psen1 c.548G>T KI/KI; Psen2−/− mice were generated by breeding Psen1 c.548G>T KI mice with Psen2−/− mice, which exhibited no detectable phenotypes (Steiner et al., 1999). The genetic background of all the mice used in this study was C57BL/6J 129 F1 hybrid. All procedures relating to animal care and treatment conformed to the Institutional and NIH guidelines.
Total RNAs were purified with TRI reagent (Sigma) according to manufacture's protocol, treated with DNase I, and reverse-transcribed in the presence of random hexamers and SuperScript® III Reverse Transcriptase (Invitrogen). PCR reactions were performed using SYBR Green PCR Master Mix with a 7500 Fast Real-Time PCR System (Applied Biosystems) using cDNA and gene-specific primers. Reactions were performed in triplicate, and threshold cycle values (Ct) were normalized to those of Gapdh. The primer pairs used in this study were as follows: 5'-TGGCCACCATCAAATCAG-3' and 5'-TCATGATGGCCGCATTCAG-3' for Psen1, 5'-TTGTCTCCTGCGACTTCA-3' and 5'-TCCACCACCCTGTTGCTGTA-3' for Gapdh. These prmer pairs were confirmed not to give rise to primer dimers during the PCR reaction. For Northern analysis, 10–20 μg of total RNAs were separated in formaldehyde agarose gels and transferred to nylon membranes (Amersham). Hybridization was performed using [α-32P] dCTP-labeled probes specific for each gene. Almost entire coding region of mouse Psen1 gene was used for Psen1-specific probe so as to detect any possible transcriptional/post-transcriptional variant from the Psen1 gene.
Dissected cortices (at 1–2 months for c.548G>T KI mice) were homogenized in RIPA buffer (50 mM Tris-Cl (pH 7.6), 150 mM NaCl, 0.5 mM EDTA, 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS, protease inhibitor cocktail (Sigma), 1mM PMSF). Equal amount (10–40 μg per lane) of proteins were separated in NuPAGE gels (Invitrogen) and transferred to nitrocellulose membranes. The membranes were blocked in 5% nonfat milk/TBS for 1 hour, and incubated with specific primary antibodies shown as below: rabbit anti-PS1 NTFs (#529591, Calbiochem), rabbit anti-PS1 CTFs (#529592, Calbiochem), rabbit anti-Pen2 (#36–7100, Zymed), rabbit anti-Nicastrin (N1660, Sigma), rabbit anti-Aph1 (PA1-2010, Thermo science), mouse anti-α-tubulin (T6199, Sigma), mouse anti-phospho-tau (PHF1 and CP13, gift of P. Davis), rabbit anti-total tau (JM, gift of A. Takashima), rabbit anti-Notch1 (V1744) (#2421, Cell Signal Tech), mouse anti-SNAP25 (MAB331, Chemicon). The membrane was then incubated with IRDye 800CW or IRDye 680-labeled secondary antibodies (LI-COR Bioscience). Signals were developed and quantified with an Odyssey Infrared Imaging System (LI-COR Bioscience).
Drug administration was performed as described previously (Contet et al., 2007). Briefly, cycloheximide (Sigma) or saline was intraperitoneally administered at one hour intervals (200 mg/kg body weight). Tissues were dissected and frozen for further molecular analysis 4 hours after the first injection.
Psen-deficient mouse embryonic fibroblasts (MEFs) were transfected with equal amounts (1ng) of expression vectors encoding wild-type or various mutant PS1 along with an N-terminally-truncated Notch1 construct (NΔE, gift of A. Goate) as described previously (Heilig et al., 2010). The amount of plasmid DNA used was derived from a dose curve analysis using 0.125, 0.25, 0.5, 1.0, 2.0, 4.0 and 8.0 ng of Psen1 wild-type cDNAs, and doses within the linear range were 0.5 (0.5×), 1.0 (1×) and 2.0 (2×) ng. A 1.7 kb mouse Psen1 coding sequence was amplified with primers 5'-GATCTCGAGTTCGAGGTCTTTAGGCAGCTTG-3' and 5'-AGTGCGGCCGCTGCTGCAGCGATGGATGTTGG-3' (the underlines are XhoI and NotI sequences, respectively), and subcloned into XhoI-NotI sites of pCI expression vector (Invitrogen). The corresponding mutations of each mutant PS1 were introduced by either subcloning or site-directed mutagenesis. Cells were harvested with RIPA buffer 24 hours after transfection, and western analysis was performed using anti-Notch1 (V1744) antibody (#2421, Cell Signal Tech). Data are normalized to α-tubulin, and three independent experiments were quantified.
γ-Secretase-mediated de novo Aβ generation was measured using a method described previously (Takahashi et al., 2003). Briefly, the cortices at 1–2 months of age were homogenized in homogenization buffer (20mM PIPES, pH 7.0, 140mM KCl, 0.25 M sucrose, 5mM EGTA) using a glass/Teflon tissue grinders. The homogenates were centrifuged at 800 g for 10 min to remove nuclei and cell debris. The postnuclear supernatants were recentrifuged at 100,000 g for 1 hour, and the resulting pellets were washed with 0.1M sodium carbonate (pH 11.4) and then centrifuged again. The membrane pellets were solubilized with 1% 3-[(3-cholamidopropyl)dimethylammonio]-2-hydroxy-1-propanesulfonate (CHAPSO) in homogenization buffer for 1 hour on ice and then centrifuged at 100,000 g for 1 h, and finally the resulting soluble fractions were saved at −80°C until use as crude γ-secretase fractions. For in vitro γ-secretase assay, the CHAPSO-soluble microsomal proteins were mixed with assay buffer (10 mM HEPES (pH 7.3), 150 mM NaCl, 1 mM EDTA, Complete protease inhibitor cocktail (Roche), 5 mM 1,10-phenanthroline, 5 mg/ml phosphoramidon, and 0.1% (w/v) phosphatidylcholine), and incubated with recombinant C100-FmH or N102-FmH (1–2μM) as a γ-secretase substrate at 37°C for 14 hours. To quantify de novo Aβ generation, samples were subjected to ELISA specific for Aβ40 and Aβ42. Specific γ-secretase activity was obtained by subtracting the values obtained from a reaction conducted in the presence of γ-secretase inhibitor (III-31C, Sigma). For Western analysis of in vitro γ-secretase assay, specific signals of cleaved substrate were normalized by the signals of SNAP25. For recombinant C100-FmH and N102-FmH, which are tagged with Flag-Myc-Histidine, bacterial strain DH5α was transformed with pTrcHis2A-C100-FmH and pTrcHis2A-N102-FmH plasmids (gifts of T. Iwatsubo and T. Tomita), respectively, and induced with 0.1mM Isopropyl β-D-1-thiogalactopyranoside (IPTG) for 3 hours. Each recombinant protein was purified with Ni2+-chelated HiTrap Chelating HP column (Amersham).
Monoclonal antibodies directed against the C terminus of Aβ40 (Covance; 11A50-B10) or Aβ42 (Covance; 12F4) were used for specific capture of Aβ species. A biotinylated monoclonal secondary antibody (Covance; 4G8) recognizing Aβ residues 17–24 was used for detection of both Aβx-40 and Aβx-42 with a reporter system of streptavidin-conjugated alkaline phosphatase (Promega) and AttoPhos reagent (Promega). Fluorescence was measured with excitation at 444 nm and emission at 555 nm by Synergy HT microplate reader (BioTek). Aβ40 and Aβ42 synthetic peptide standards (BioPeptide Inc) were included in each analysis for quantifying Aβ levels, which were expressed as the concentration in pM of Aβx-40 or Aβx-42. For endogenous mouse Aβ ELISA, mouse cortices were homogenized in 0.2% diethylamine and centrifuged at 100,000 g. The resulting supernatants were neutralized with Tris-Cl (pH 6.8), and directly used for Aβ measurement. Monoclonal antibodies were used for capture of Aβ40 (266) and Aβ42 (21F12), and biotinylated monoclonal secondary antibodies were used for detection of Aβ40 (2G3) and Aβ42 (266). Endogenous Aβ concentration was calculated to divide total Aβ amount by total protein amount in the cortex fraction.
The Morris water maze is a circular pool 160 cm in diameter. Mice were housed in a standard 12 hr light-dark cycle. Prior to testing, the experimenter handles each mouse for 5 minutes a day for 5 days. During the hidden platform training, the platform (10 cm in diameter) was kept submerged under water and maintained in the same position. Each mouse was given four trials daily with a maximum duration of 90s separated by a minimum of 15 min. If the mouse did not find the hidden platform, it was guided to the platform and allowed to remain on it for 15s. The swimming of the mice was monitored using an automated tracking system (HVS Image). A group of mice at the age of 6–8 (Psen1 c.548G>T KI/KI;Psen2 −/− and Psen1 +/+;Psen2 −/−) or 18–20 (Psen1 c.548G>T KI/KI and Psen1 +/+) months were trained in the hidden platform task for 12 days. Twenty-four hours after the last training (day 12), the hidden platform was removed and a 90 s probe trial was performed to test a spatial reference memory. The mice were released from all four quadrants in a pseudorandom manner during the training and the probe trial as possible. Finally, visible cue task with four trials was performed to verify visual and swimming ability using the platform marked by a yellow object. The experimenters were blind to the genotypes of the mice.
Brains were perfused with PBS, fixed in 4% paraformaldehyde for 3 hours at 4°C, and processed for paraffin embedding. Paraffin-embedded sagittal sections were serially cut at 10μm. Sections in every 40 slides were deparaffinized, dehydrated, and stained with 0.5% Cresyl Violet (Sigma), and analyzed for brain volumes by BioQuant image analysis software. For immunohistochemistry, paraffin-embedded brain sections were deparaffinized, alcohol dehydrated, and immunostained with monoclonal antibodies raised against phosphorylated (pSer396/pSer404) tau (PHF1, gift of P. Davis). After specific signals were developed by Vectastain Elite ABC kit and DAB peroxidase substrate, the sections were lightly counterstained with haematoxylin. The signals were analyzed by BX50 microscope system (Olympus).
Statistical analyses were performed using one-way ANOVA or two-tailed unpaired Student's t-test for all the comparisons in the behavioral and biochemical results. A value of P < 0.05 was considered significant. All the data were described as mean ± SEM.
To investigate the pathogenic mechanism underlying the Psen1 c.548G>T mutation, we generated a KI mouse in which the c.548G>T mutation was introduced into Psen1 exon 6 by homologous recombination. The targeting vector includes the 5' homologous region (2.49 kb), floxed PGK-neo cassette and the 3' homologous region (3.15 kb), in which the c.548G>T mutation was introduced into the last nucleotide of exon 6 (Fig. 1A, B). The embryonic stem (ES) cells carrying the proper homologous recombination events in the 5' and 3' homologous regions were confirmed by Southern analysis using the 5' and 3' external probes, respectively (Fig. 1B, data not shown), and injected into mouse blastocysts to generate chimeric mice, which were then used to generate heterozygous mice carrying the targeted allele (targeted/+). Southern analysis of tail genomic DNAs of these mice confirmed that homologous recombination and germline transmission occurred correctly, as demonstrated by the presence of two bands representing the wild-type allele (11.0 kb for both 5' and 3' external probes) and the targeted allele (5.5 kb for the 5' external probe, 8.5 kb for the 3' external probe) (Fig. 1C). The floxed PGK-neo selection cassette was removed by crossing the F1 mice transmitted with the targeted allele to Cre-expressing transgenic mice so as to avoid its possible transcriptional interference on Psen1 expression (Fig. 1B). Southern analysis of tail genomic DNAs from the resulting mice confirmed the deletion of the floxed PGK-neo selection cassette using the 3' external probe (Fig. 1C). The targeted allele following the removal of the floxed PGK-neo cassette is termed the KI allele (Fig. 1B, C). Additional Southern analysis using the neo probe further confirmed the lack of the floxed PGK-neo cassette as well as the absence of random insertion of the targeting vector in the KI mice (data not shown). The correct introduction of the c.548G>T mutation was also confirmed by sequencing (Fig. 1D).
The c.548G>T KI/KI homozygous mice were born in Mendelian ratio and were fertile, and adult KI/KI mice are grossly normal in appearance compared to littermate controls. Histological analysis of brain morphology shows that KI/KI brains were indistinguishable from littermate controls (Fig. 1E). Using stereological methods, we measured cortical volume in Nissl-stained series sagittal sections and found similar cortical volumes between KI/KI and littermate controls even at 20 months of age (Fig. 1E). Western analysis of GFAP, which is elevated accompanying astrogliosis (Beglopoulos et al., 2004; Saura et al., 2004), indicated no increases in the cerebral cortex of KI/KI mice at 12–20 months of age (Fig. 1F). These results suggest that Psen1 c.548G>T KI/KI mice don't develop age-dependent neurodegeneration.
Since the c.548G>T mutation resides on the splice junction of exon 6 and intron 6, which likely disrupts normal splicing of its transcripts, we evaluated Psen1 mRNA expression in c.548G>T KI mice. Northern analysis revealed that levels of Psen1 mRNAs are significantly reduced in the cerebral cortex of KI/KI mice throughout their life (from postnatal day 1 to 10 months) (Fig. 2A). Quantitative qRT-PCR also confirmed the reduction of Psen1 mRNAs in c.548G>T KI mice in a KI allele dose dependent manner, with a ~40% reduction in KI/KI mice (Fig. 2B). Western analysis disclosed ~30% reduction in levels of PS1 N-terminal fragments (NTFs) and C-terminal fragments (CTFs) in total cortical lysates (Fig. 2C). Given that levels of Psen1 mRNAs correlated with levels of PS1 proteins in Psen1−/− and Psen1+/− mice (Shen et al., 1997), the discrepancy in the reduction of Psen1 mRNAs (~40%) and proteins (~30%) in c.548G>T KI/KI mice may reflect a mechanism suggested in a previous report (Lee et al., 1997), in which mutant PS1 tends to accumulate in transgenic mouse brains. In contrast, the protein levels of other γ-secretase complex components, nicastrin (NCT), Aph-1, and Pen-2, were similar among the three genotypic groups (Fig. 2C).
We next investigated the mechanism underlying reduced levels of Psen1 mRNAs in the brain of KI/KI mice. The location of the G to T transversion responsible for the c.548G>T mutation suggested that splicing of Psen1 transcripts might be disturbed in KI/KI mice (Fig. 1A), because proper splicing requires the exon-intron junction to conform to the consensus sequences (Cartegni et al., 2002). Indeed, RT-PCR using primers in exons 4 and 10 followed by sequencing showed that in addition to normal splice products containing all exons 4–10, two aberrant splice products lacking either exon 6 or exons 6 and 7 were produced in c.548G>T KI/KI brains (Fig. 3A). Since the skipping of exon 6 or both exons 6 and 7 results in frame-shift and use of downstream premature termination codon that are normally out of frame, we then evaluated whether these abnormal transcripts are degraded by nonsense-mediated decay (NMD) mechanisms, which is a surveillance system to prevent production of truncated proteins (Cartegni et al., 2002). We used cycloheximide, a potent inhibitor of protein synthesis and nonsense-mediated decay, to determine whether blockade of NMD would restore the levels of aberrant splice products that were normally degraded by NMD. Indeed, cycloheximide treatment enhanced the stability of the aberrantly spliced transcripts and drastically increased levels of the splice product lacking exons 6 and 7 in KI/KI brains (Fig. 3B). Quantitative RT-PCR analysis also showed that cycloheximide treatment fully rescued the reduction of Psen1 mRNA levels in KI/KI brains (Fig. 3C). These results reveal that in addition to the Glycine to Valine conversion at the amino acid residue 183 (p.183G>V), the c.548G>T mutation disrupts proper splicing of a portion of the Psen1 transcripts and results in generation of aberrantly spliced products harboring premature termination codons, which were rapidly degraded by NMD mechanisms, leading to a consequent ~40% decrease in the expression of full length Psen1 mRNAs in KI/KI brains.
PS1 is ubiquitously expressed in many tissues (Lee et al., 1996), and reduced Psen dosage has been associated with skin disorders in humans and mice (Xia et al., 2001; Tournoy et al., 2004; Kelleher and Shen, 2010; Wang et al., 2010; Pink et al., 2011). But in FTD patients, the c.548G>T mutation gives rise to a brain-specific phenotype with no clinical remarks in other tissues (Dermaut et al., 2004). We therefore examined whether the c.548G>T mutation affects Psen1 splicing and expression in non-neural tissues. Interestingly, RT-PCR showed no evidence of aberrant splicing in the liver, lung and skin of KI/KI mice, whereas the aberrantly spliced mRNAs were readily detected in the neocortex, hippocampus and cerebellum (Fig. 4A). As a result, levels of Psen1 mRNAs are unaltered in the liver, lung and skin, as shown by quantitative RT-PCR (Fig. 4B). Additionally, we also examined expression levels of Psen1 mRNAs in splenocytes, which contain broad spectrums of lymphocyte cell lineages, because Notch, a substrate of γ-secretase, is involved in regulation of lymphocyte cell development (Pui et al., 1999; Radtke et al., 1999). Northern analysis showed that expression levels of Psen1 mRNAs are not changed in splenocytes in KI/KI mice (data not shown). Moreover, cycloheximide treatment results in an accumulation of aberrantly spliced Psen1 transcripts in the brain but not in the liver of c.548G>T KI/KI mice (Fig. 4C). These results demonstrate that the c.548G>T mutation causes aberrant splicing and subsequent reduction of Psen1 expression only in the brain but not in non-neural tissues.
To determine the effect of the c.548G>T mutation on γ-secretase activity, we performed in vitro γ-secretase assays using detergent-solubilized cerebral cortical fractions as the sensitive, direct method (Takahashi et al., 2003). Using CHAPSO-solubilized fractions from the cerebral cortex of c.548G>T KI/KI, KI/+ and wild-type littermate control mice at 2–3 months of age, we found that γ-secretase activity is reduced in KI/+ mice and further decreased in KI/KI mice, as measured by de novo production of Notch intracellular domain (NICD) from recombinant N102-FmH (substrates of NICD) (Fig. 5A). Production of NICD was abolished by a specific γ-secretase inhibitor III-31C (Fig. 5A).
APP is the other well-established physiological substrate of γ-secretase and γ-secretase-mediated cleavages give rise to Aβ40 and Aβ42 peptides (De Strooper et al., 1998). We then performed the same in vitro γ-secretase assay using CHAPSO-solubilized cortical fractions of c.548G>T KI/KI, KI/+ and wild-type mice, and found that de novo generation of Aβx-40 and Aβx-42 is drastically reduced in KI mice depending on the KI allele dosage (Fig. 5B). As a control, we also used cortices of Psen1 conditional KO (cKO) mice, and similar to our prior finding (Yu et al., 2001), de novo generation of Aβx-40 and Aβx-42 is robustly reduced in Psen1 cKO mice (Fig. 5B). In addition, we confirmed a decrease of de novo generation of total Aβ by Western blot (data not shown). We further performed sandwitch ELISA to measure steady state levels of endogenous Aβ peptides. Interestingly, measurement of the peptides showed decreased levels of Aβ40 but unchanged levels of Aβ42 in the cerebral cortex of c.548G>T KI/KI mice at 12 months of age (Fig. 5C). The difference between de novo Aβ production and steady state Aβ levels is likely due to the effect of Aβ turnover in vivo (Wang et al., 2006).
Although the reduction of γ-secretase activities in c.548G>T KI mice is consistent with decreased Psen1 mRNA levels in these mice, the c.548G>T mutation also results in a substitution of Glycine with Valine at amino acid residue 183. To determine whether the p.183G>V conversion itself alters γ-secretase activity, we transfected vectors expressing either wild-type at varying amounts (1×, 2×, 0.5×) or various mutant Psen1 cDNAs (1×) and truncated Notch1 (NΔE) into Psen-deficient mouse embryonic fibroblasts (MEFs) (Herreman et al., 2000). One day after the transfection, we collected cell lysates and performed Western analysis. Wild-type and p.183G>V PS1 displayed similar levels of γ-secretase activity, as measured by production of NICD, whereas PS1 bearing an p.257D>A substitution, which abolishes γ-secretase activity, was unable to produce detectable NICD (Fig. 5D). The Psen1 mutant cDNA lacking exons 6 and 7 was similarly devoid of γ-secretase activity (Fig. 5D). These results suggest that full-length PS1 harboring the p.183G>V alteration has normal γ-secretase activity, thus, this change is unlikely to be pathogenic. Collectively, these results suggest that the pathogenic effect of the c.548G>T mutation is a consequence of the brain-specific reduction in Psen1 mRNA expression.
In addition to cerebral atrophy, FTD is sometimes associated neuropathologically with tau pathology (Neary et al., 2005). Neuropathological analysis of the patient carrying the PSEN1 c.548G>T mutation revealed Pick's body and phosphorylated tau staining, in addition to severe frontotemporal atrophy and clear neuronal loss (Dermaut et al., 2004). Therefore we next examined whether levels of tau and phosphorylated status are elevated in the cerebral cortex of KI/KI mice. Western analysis using rabbit polyclonal antibody for total tau showed similar levels of tau in the hippocampus and the neocortex between KI/KI and wild-type littermate mice (Fig. 6A and data not shown). Using monoclonal antibodies, PHF1 (specific for phosphorylated Ser396/Ser404) and CP13 (specific for phosphorylated Ser202/Thr205), we found no significant changes phospho-tau in the neocortex and hippocampus of KI/KI mice (Fig. 6A). Immunohistochemical analysis using PHF1 antibodies further confirmed similar levels of phospho-tau in the neocortex and hippocampal areas CA1 and CA3 regions of KI/KI mice compared to littermate controls at 20 months of age (Fig. 6B). These results show that the Psen1 c.548G>T mutation does not cause tau pathology in mice, which may be due to the subtle effect of the mutation and the short lifespan of mice.
We previously reported that loss of presenilin function in the postnatal cerebral cortex impairs learning and memory in a gene dosage-dependent manner (Yu et al., 2001; Saura et al., 2004). Whereas Psen2−/− mice exhibited normal learning and memory, conditional inactivation of Psen1 and Psen1/2 in postnatal cerebral cortex caused mild and severe memory deficits, respectively. To determine whether the c.548G>T mutation impairs presenilin function in cognition, we tested hippocampus-dependent spatial learning and memroy using the Morris water maze task. Psen1 c.548G>T KI/KI and wild-type littermate control mice were given 4 trials a day for 12 days, and they performed similarly during the 12 day learning phase (e.g. latency and swim speed in Fig. 7A, B) and in the probe trial 24 hours after the final training session (e.g. quadrant occupancy in Fig. 7C). Psen1 c.548G>T KI/KI also did not show any deficit in the visible cue task (Fig. 7D). We next examined Psen1 c.548G>T KI/KI and littermate control mice in the Psen2-null background, because Psen1 cKO mice exhibit more severe learning and memory deficits in the absence of PS2 compensation (Saura et al., 2004). During the 12-day training phase, ANOVA analysis did not show a significant genotype effect in escape latency (Fig. 7E, F=0.709; df=1,17; P=0.4114), path length (F=0.032; df=1,17; P=0.8601) and swim speed (Fig. 7F, F=3.549 ; df=1,17; P=0.0768), but Psen1 c.548G>T KI/KI; Psen2−/− mice exhibited significantly lower quadrant occupancy during the probe trial administered 24 hours after the last training session (Fig. 7G, p<0.05), suggesting mild spatial memory impairment. The Psen1 c.548G>T KI/KI; Psen2−/− mice performed normally in the visible cue task (Fig. 7H). These results suggest that the Psen1 c.548G>T mutation causes a functional impairment of presenilin activity in cognition likely due to reduction of Psen1 expression.
Presenilin comprises the catalytic component of the aspartate protease complex, γ-secretase, which plays essential roles during embryonic development (Shen et al., 1997; Wong et al., 1997; Donoviel et al., 1999; Li et al., 2000). The γ-secretase complex, which also includes Nicastrin, Aph-1, and Pen-2 (Yu et al., 2000; Francis et al., 2002; Goutte et al., 2002), is involved in the intramembrane cleavage of type I transmembrane proteins such as amyloid precursor protein (APP) and Notch (De Strooper et al., 1998; De Strooper et al., 1999). Presenilin is broadly expressed in the brain but mediate unique functions in specific cell types (Lee et al., 1996; Handler et al., 2000; Saura et al., 2004; Wines-Samuelson et al., 2005; Kim and Shen, 2008; Zhang et al., 2009; Wines-Samuelson et al., 2010). Importantly, we previously showed that selective inactivation of PS in the adult cerebral cortex causes progressive memory impairment followed by age-dependent neurodegeneration in the absence of increases of β-amyloid peptides (Beglopoulos et al., 2004; Saura et al., 2004; Wines-Samuelson et al., 2010). Although these studies support the view that loss of PS function contributes to neurodegeneration and dementia, how PSEN mutations affect its function and lead to AD and FTD remains to be determined.
In this study, we generated a novel KI mouse model carrying the FTD-associated PSEN1 c.548G>T mutation to investigate the underlying pathogenic mechanism and to explore the molecular link between the roles of PSEN mutations in AD and FTD. Unexpectedly, we discovered that the c.548G>T mutation reduces expression of full-length Psen1 mRNAs as a result of aberrant splicing and NMD-mediated degradation of improper splice products (Figures 2, ,3),3), and that the corresponding p.183G>V change in protein sequence does not affect PS function in γ-secretase activity (Figure 5). Consistent with this reduction in Psen1 expression, c.548G>T KI/KI brains exhibit reduced γ-secretase activity and mild learning and memory deficits (Figures 5 and and7).7). Surprisingly, the aberrant splicing and decreased Psen1 expression elicited by the c.548G>T mutation are brain-specific, illuminating a mechanism by which Psen1 mutations can produce brain-specific phenotypes in FTD patients (Figure 4). Collectively, our genetic study shows that the FTD-associated PSEN1 c.548G>T mutation causes a net loss of presenilin activity due to reduced Psen1 mRNA expression (Figure 8).
Interestingly, aberrant splicing of PSEN1 transcripts and abnormal PS1 protein expression has previously been reported in cases of sporadic FTD (Evin et al., 2002). Moreover, PSEN1 mutations identified in other FTD pedigrees also reside at exon-intron boundaries, possibly leading to splicing defects (Raux et al., 2000; Borroni et al., 2011). Thus, failure of proper PSEN1 mRNA splicing and consequent reduction of PS1 expression could represent a common mechanism underlying both sporadic and familial FTD. Supporting the plausibility of such a mechanism in FTD pathogenesis, it has been estimated that approximately 15% of all point mutations causing human genetic disease result in an mRNA splicing defect (Krawczak et al., 1992; Liu et al., 2001).
Another intriguing discovery of our study is the brain specificity of the splicing defect caused by the c.548G>T mutation, sparing other tissues from detrimental effects associated with reduced PSEN1 expression. Recent studies have shown an association between haploinsufficiency of γ-secretase component proteins and familial acne inversa (Kelleher and Shen, 2010; Wang et al., 2010; Li et al., 2011; Pink et al., 2011). However, our molecular analysis of c.548G>T KI mice showed that levels of Psen1 mRNAs are only reduced in the brain (Figure 4). Further analysis revealed that aberrant splicing and subsequent degradation of transcripts from the c.548G>T KI allele occur only in the brain but not in other tissues. The detailed molecular mechanism by which aberrant splice products are produced in a brain-specific manner remains to be determined. One possible explanation could be the existence of brain-specific alternative splicing factors (Dredge et al., 2001). A brain-restricted expression of a specific splicing factor(s) might render the splicing machinery less tolerant of the G>T transversion at the splice donor site, leading to skipping of this exon only in the brain.
While the effect of the c.548G>T mutation on the overall level of Psen1 mRNAs is modest, its effect on γ-secretase activity measured by de novo production of NICD or Aβ peptides in KI brains is surprisingly robust (Figure 5). For example, the production of Aβ40 peptides in c.548G>T KI/KI brains is reduced by ~70%. These results indicate that γ-secretase activity is very sensitive to changes in presenilin dosage, suggesting that small reductions of presenilin expression may be sufficient to produce substantial deficits in its essential functions (e.g. at the synapse). However, in vivo steady state levels of Aβ peptides are less affected (Figure 5), suggesting that a compensatory reduction in Aβ turnover may accompany the reduced Aβ production in KI mice. Consistent with our earlier findings (Yu et al., 2001; Saura et al., 2004), we found that Psen1 c.548G>T KI/KI mice in the Psen2-null background exhibit mild deficits in spatial learning and memory (Fig. 7). Not surprisingly, the spatial memory deficits exhibited by Psen1 c.548G>T KI/KI; Psen2−/− mice was subtler than those of Psen1 cKO mice, in which PS1 is completely inactivated in excitatory neurons of the adult cerebral cortex, whereas Psen cDKO mice exhibit more dramatic memory impairment (Yu et al., 2001; Saura et al., 2004).
It has been long debated whether PSEN mutations cause a toxic gain of function or a loss of essential functions normally carried out by presenilin (Shen and Kelleher, 2007). PSEN mutations associated with familial AD often lead to selective increases of the longer and more amyloidogenic Aβ42 peptides, which has been taken as evidence for the toxic gain-of-function mechanism. However, shortly after the identification of the PSEN mutations in FAD, such mutations were found to reduce its biological function and γ-secretase activity in invertebrate models and cell culture systems (Levitan et al., 1996; Song et al., 1999; Seidner et al., 2006). More recently, a clinical PSEN1 mutation conferring a complete loss of PS1 function and γ-secretase activity has been described (Heilig et al., 2010). These findings are consistent with our prior work showing that complete inactivation of presenilin function in the adult mouse brain causes dementia and progressive neurodegeneration, two key features common to AD and FTD (Saura et al., 2004; Wines-Samuelson et al., 2010). Thus, impairment of essential PS functions in neuronal survival and memory may be a common property of PSEN mutations in both of these neurodegenerative dementias.
Our findings demonstrate that the FTD-associated c.548G>T mutation decreases Psen1 mRNA expression, leading to a partial loss of presenilin and γ-secretase function. However, the amino acid substitution derived from the c.548G>T mutation, p.183G>V, does not affect γ-secretase activity. Thus, the net effect of the c.548G>T mutation is limited to the reduction of Psen1 mRNA and protein expression, indicating that a pure loss of PSEN expression is pathogenic. This loss of PS expression without altered PS protein activity offers a possible mechanism to account for the development of neurodegeneration and dementia in the absence of Aβ deposition in FTD. It is presently unclear whether other PSEN mutations that have been associated with FTD also affect PS expression, although the localization of some FTD-associated PSEN mutations at splice junctions (e.g. c.1129A>T, c.338T>C) raises the possibility that similar mechanisms may be at play (Ikeda et al., 1996; Raux et al., 2000; Binetti et al., 2003; Portet et al., 2003; Rippon et al., 2003; Halliday et al., 2005; Zekanowski et al., 2006; Bernardi et al., 2009; de Bot et al., 2009; Marcon et al., 2009; Gallo et al., 2010; Borroni et al., 2011). Future studies will be needed to determine the effects on presenilin function and γ-secretase activity of other FTD-associated PSEN mutations, which include mutations situated at the splice consensus sites (PSEN1 c.1129A>T [p.377R>W], c.338T>C [p.113L>P]) and mutations whose localization suggests simple missense substitution (e.g. PSEN1 p.139M>V, p.146M>L, p.226L>F, p. 233M>L, p.260A>V, p.412V>I, p.424L>H, p.424L>R; PSEN2 p.62R>H, p.122T>R, p.231Y>C, p.239M>V). These studies may help us understand why some of the same PSEN1 mutations are associated with patients who were initially diagnosed clinically as AD or FTD (Ikeda et al., 1996; Portet et al., 2003; Rippon et al., 2003; Halliday et al., 2005; Zekanowski et al., 2006). Our findings imply that transgenic approaches in cell culture or mice may not provide a complete picture of the impact of PSEN mutations on PS function and γ-secretase activity; rather, analysis of PSEN mutations in the context of the genomic locus will be important to evaluate potential effects on mRNA splicing and post-transcriptional regulation of PSEN expression. Whether the pure loss of PSEN expression identified in this study constitutes a common pathogenic mechanism by which PSEN mutations cause FTD remains to be determined.
We would like to thank H. Zhao and X. Zou for breeding and genotyping the mice, and other lab members for helpful discussions. We also thank T. Iwatsubo and T. Tomita for pTrcHis2A-C100-FmH and pTrcHis2A-N102-FmH plasmids, B. DeStrooper for Psen DKO MEFs, P. Davis for phospho-specific tau antibodies (PHF1 and CP13), A. Takashima for total tau antibody (JM). This work was supported by grants from the National Institutes of Health (R01NS041783, RC2AG036614 to J.S.) and grants from the Alzheimer's Association (to R.J.K., J.S.).
Conflict of interest: The authors declare no conflict of interest.
Author contributions: H.W., R.J.K., and J.S. designed research and wrote the paper; H.W., D.X., and T.K. performed experiments; H.W., R.J.K., and J.S. analyzed data.