The detailed information for the antibodies used in this work is listed in Table S1 in Supplementary Material. R_p-adenosine 3′, 5′-cyclic monophosphorothioate triethyl ammonium salt (R_p-cAMPS, a specific inhibitor of PKA) and okadaic acid (OA) were purchased from Sigma. To knockdown I2PP2A in cells, shRNA oligo sequences were synthesized as follow: 5′-AGCTTGGATGAAGGTGAAGAAGATTTCAAGAGAATCTTCTTCACCTTCATCCTTTTTC-3′, 5′-TCGAGAAAAAGGATGAAGGTGAAGAAGATTCTCTTGAAATCTTCTTCACCTTCATCCA-3′. As control, we used non-functional I2PP2A-derived sequences: 5′-AGCTTTGAGAGTGGTGATCCATCTTTCAAGAGAAGATGGATCACCACTCTCATTTTTC-3′, 5′-TCGAGAAAAATGAGAGTGGTGATCCATCTTCTCTTGAAAGATGGATCACCACTCTCAA-3′ (ten Klooster et al., 2007; Liu et al., 2008). All were purchased as 63-nt ssDNA oligomers composed of both forward and reverse sequences with 9-bp loop structures (Brummelkamp et al., 2002) and 3′ Xho1 I and 5′ HindIII self-inactivating overhangs. Sense and antisense oligomers (both at 20 μM) were incubated in annealing buffer for 3 min at 90°C as described (Elbashir et al., 2001), then the temperature was lowered in 2°C/min increments until 5°C above their respective T_m and then dropped to 4°C at maximum ramp rates. pSUPER, a mammalian expression vector that directs the synthesis of siRNAs (Brummelkamp et al., 2002) was digested with both Xhol I and HindIII. Annealing shRNA was cloned into Xhol I and HindIII-digested pSUPER (pSUP): pSUPER - siI2PP2ApSUP - siI2PP2A and pSUPER-siCon (pSUP-siC).

Vector plasmids were constructed for the production of third-generation lentivirus-expressing siRNA for I2PP2A. Fortunately, the siRNA target for human I2PP2A also paired with the sequence of mouse I2PP2A and knockdown I2PP2A level in N2a cells. All vectors contained the eGFP coding sequence located in the middle of the lentiviral vector. This sequence is driven by a cytomegalovirus (CMV) promoter and terminates using the polyadenylation signal in the 3′ long terminal repeat (LTR). Downstream of the eGFP is a woodchuck hepatitis virus regulatory element (WPRE) that enhances the expression of the transgene. Recombinant lentiviruses were produced by transient transfection in HEK293T cells using the calcium phosphate transfection method, as described previously (Naldini et al., 1996). The infectious lentiviruses were harvested at 48 and 72 h post-transfection and filtered through 0.22-μm-pore cellulose acetate filter. The infectious lentiviruses were concentrated by ultracentrifugation (2 h at 50,000 × g) and subsequently purified by ultracentrifugation on a 20% sucrose gradient (2 h at 46,000 × g) as described (Naldini et al., 1996). Vector concentrations were analyzed using an immunocapture p24-gag ELISA (Alliance; DuPont-NEN; Naldini et al., 1996) and by flow cytometry quantification of eGFP-positive transduced cells, as described previously (Marr et al., 2003).

The human tau transgenic mice (htau, ~11-month-old) [STOCK Mapt^tm1(EGFP)Klt Tg(MAPT)8cPdav/J, Jackson Lab], which express six isoforms of tau and show an age-dependent development tau pathology and impairments of cognitive and synaptic functions (Polydoro et al., 2009) were used for the study. For brain injections, the mice were positioned in a stereotaxic instrument and 2 μl Lenti - siI2PP2A or Lenti - ssiI2PP2A were injected into the hippocampus (AP – 2.0, ML – 1.5, DV – 2.0; Kaspar et al., 2002) at a rate of 0.50 μl/min. The syringe was left in the place for ~3 min before being slowly withdrawn from the brain. After 4 weeks, the mice were sacrificed, and the hippocampi were quickly removed out and homogenized on ice in lysis buffer [50 mM Tris–HCl (pH 7.5), 150 mM NaCl, 1% (v/v) Triton X-100, 1% (w/v) deoxycholate, 0.1% (w/v) SDS, 10 mM NaF, 1 mM Na₃VO₄, and 2 μg/ml each of aprotinin, leupeptin, and pepstatin A], then brain extracts stored in −80°C. All mice were kept at 24 ± 2°C on daily 12 h light–dark cycles with ad libitum access to food and water. The animal experiments were carried out according to the “Policies on the Use of Animals and Humans in Neuroscience Research” approved by the Society for Neuroscience in 1995, and also approved by Institutional Animal Care and Use Committee at Tongji Medical College, Huazhong University of Science and Technology.

Four weeks after the brain infusion of the lentiviral vectors, the step-down avoidance test was performed by following a previous procedure (Zarrindast et al., 2006). Briefly, the apparatus consisted of an open field gray Plexiglas box (40 cm × 40 cm) with a steel rod floor. The Plexiglas platform (4 cm × 4 cm × 4 cm) was set in the center of the grid floor. Intermittent electric shocks (20 mA, 50 Hz) were delivered to the grid floor by an isolated stimulator. On the first day, each mouse was gently placed on the platform. When the mouse stepped down from the platform and placed all its paws on the grid floor, an intermittent electric shock was delivered for 3 s. Responsiveness to the punishment in the training test was assessed by the animal’s vocalization, only those mice that vocalized touching the grid with the four paws were used for the retention test in order to exclude the mice with a different pain threshold. Two hours [short-term memory (STM)] or 24 h [(long-memory (LTM)] after training, each mouse was placed on the platform again. The first time spent before stepping down onto the grid (latency period) and frequency (number of errors) stepping down the platform were measured, considering 300 s as the upper cut-off, during the training and retention tests.

The human embryonic kidney 293 cells or mouse N2a neuroblastoma cells stably expressing the longest human tau (tau441) cDNA (HEK293/tau or N2a/tau) were cultured in DMEM supplemented with 10% fetal bovine serum (FBS). The cells were maintained at 37°C in 5% CO₂. The cells were plated onto six-well plates overnight and pSUP, pSUP-siC, or pSUP - siI2PP2A plasmid was transfected the next day using Lipofectamine 2000 according to the manufacturer’s instruction.

The cells were transfected with pSUP, pSUP-siC, or pSUP - siI2PP2A plasmids. After 24 h, the cell lysate was prepared by adding lysis buffer [20 mM MOPS, 50 mM β-glycerophosphate, 50 mM sodium fluoride, 1 mM sodium vanadate, 5 mM EGTA, 2 mM EDTA, 1% NP40, 1 mM dithiothreitol (DTT), 1 mM benzamidine, 1 mM phenylmethanesulfonyl fluoride (PMSF), and 10 μg/ml leupeptin and aprotinin, pH 7.2]. The activity of PKA in the extract was assayed using a PKA kinase activity assay kit (Assay Designs, Inc.,) according to the manufacturer’s protocol.

The activity of PP2A was assayed using a serine/threonine phosphatase assay kit (Promega, MA, USA). The assay was based on determining the amount of free phosphate generated in the reaction by measuring the absorbance of a molybdate malachite green–phosphate complex. Cell extracts were prepared as follows: cells were rinsed twice with ice-cold phosphate-buffered saline and then scraped into 1 ml of ice-cold phosphatase buffer [50 mM Tris, pH 7.0, 0.1 mM ethylenediaminetetraacetic acid/ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid, 1 mM DTT, 0.1% (v/v) Triton X-100, benzamide, leupeptin, 4-(2-aminoethyl) benzene-sulfonyl fluoride⋅HCl (AEBSF), and pepstatin A]. The resulting cell suspension was lysed by brief sonication and cell debris were pelleted at 15,000 × g for 30 min. Free intracellular phosphate and ATP were removed from this resulting supernatant in a spin column containing Sephadex G-25 according to the supplier’s instructions. The sample (10 μg) was incubated on a 96-well plate together with a peptide substrate RRA(pT)VA and PP2A-specific reaction buffer (50 mM imidazole, pH 7.2, 0.2 mM EGTA, 0.02% β-mercaptoethanol, 0.1 mg/ml BSA) for 30 min at 30°C. After incubation, the molybdate complex dye was added and incubated for an additional 30 min at room temperature for color development. The level of molybdate malachite green–phosphate complex formed was monitored at 630 nm.

Golgi staining was performed according to methods as followed (Woolley and McEwen, 1992). The mice (n = 3 per group) were killed by overdose of chloral hydrate, and perfused through the aorta with 200 ml 0.9% NaCl containing 0.5% sodium nitrite followed by 500 ml 0.9% NaCl containing 5% formaldehyde. Then, the brain was fixed in situ by perfusion of Golgi fixative (0.9% NaCl, 5% formaldehyde, 5% potassium dichromate, 5% chloral hydrate) in the dark. The brain was removed and processed for rapid Golgi staining in the dark. Briefly, the brain was post-fixed for 3 days in the same Golgi fixative, and impregnated with 1.0% aqueous silver nitrate solution for 3 days. Coronal brain sections of hippocampal tissue were cut at 35 μm using a vibratome (VT1000S, Leica, Germany). The images were observed by using a microscope (Olympus BX60, Tokyo, Japan). Neurons in the CA3 region which fulfill the following criteria were selected for the analysis; (i) the cell type must be identifiable, (ii) image resolution should be sufficient to visually distinguish dendritic spine formation from variably contrasting background, and (iii) completeness of Golgi impregnation of all dendrites. Subjective bias in spine counting was eliminated by prior coding of slides.

To analyze the dendritic morphology, Z-stacks (step size 1 μm) from five to seven cells were generated using a confocal microscope (LSM510, Zeiss) in bright-field mode (20× objective) and reconstructed in ImagePro in combination with the NeuroDraw toolbox for each animal. Total dendritic length and number of branch points were analyzed using NeuroExplorer software (MBF Bioscience, Williston, VT, USA).

To acquire images for spine analysis, the dendritic segments were imaged under bright-field illumination on a Zeiss Axio imager microscope with a 63× oil immersion objective, and spine morphology was analyzed according to a previously reported method (Magarinos et al., 2011), which does not assess spine density in a three dimensional manner but focuses on spines paralleled to the plane of section. Although the method may underestimate the total number of spines, it facilitates a direct comparison of treatment groups when they are analyzed in an identical manner. ImageJ software was used to calculate linear spine density (Spires-Jones et al., 2011), which was presented as the number of spines per 10 mm of dendrite length. The spine density was determined in two segments of dendrites at a distance of 90–110 μm (proximal) and 190–210 μm (distal) from the soma. From each animal, four neurons were selected from one slide, accounting for 36 neurons/per animal.

Total RNA was isolated by using Trizol™(Invitrogen Life Technologies, Carlsbad, CA, USA) according to manufacturer’s instruction. Then total RNA (3 μg in 25 μl) was reversely transcribed and the produced cDNA (1 μl) was used to detect the transcripts. Real-time polymerase chain reaction (PCR) to determine gene copy number was performed using the Rotor-Gene 3000 Real-Time PCR Detection System (Corbett Research, Sydney, NSW, Australia) with the SYBR^® Premix Ex Taq™[Takara Biotechnology (Dalian) Co., Ltd., Dalian, China]. The expression level of GADPH housekeeping gene was used for normalization of GSK-3β mRNA expression level. Forward primer 5′-ACGCTCCCTGTGATTTATG-3′ and reverse primer 5′-CAAGAGGTTCTGCGGTTTA-3′ for GSK-3β; forward primer 5′-CTTCAACTCTGGTCAAATAATGCA-3′ and reverse primer 5′-GAACAAAAATATAACAAACTCCGC-3′ for I2PP2A; forward primer 5′-GAAGGTGAAGGTCGGAGTC-3′ and reverse primer 5′-GAAGATGGTGATGGGATTTC-3′ for GADPH.

Western blotting was performed according to methods established in our laboratory. Briefly, the cell homogenates or brain extracts were mixed with sample buffer containing 50 mM Tris–HCl (pH 7.6), 2% SDS, 10% glycerol, 10 mM DTT, and 0.2% bromophenol blue and boiled for 5 min. The proteins were separated by 10% SDS/PAGE and transferred to PVDF membrane. Immunostaining was visualized with a chemiluminescent substrate kit and CL-XPosure Film and quantitatively analyzed by digital science 1D software (Eastman Kodak, Rochester, NY, USA). Band intensity was measured as the sum optical density and expressed as a level relative to each control. The phosphorylated levels of tau were normalized relative to the total tau.

The transgenic mice were sacrificed by overdose chloral hydrate (1 g/kg) after injection of lentiviral vectors for about 1 month, and perfused through aorta with 100 ml 0.9% NaCl followed by 400 ml phosphate buffer containing 4% paraformaldehyde. Brains were removed and post-fixed in perfusate overnight and then cut into sections (20 μm) with a vibratome (Leica, Nussloch, Germany; S100, TPI). Sections were incubated at 4°C overnight with primary antibodies (see Table S1 in Supplementary Material and Figure legends). The images were observed using a laser scanning confocal microscope (Olympus FV500, Tokyo, Japan).

For cell studies, cells were cultured on coverslips and fixed with 4% paraformaldehyde for 1.5 h at 4°C and then incubated for 12–36 h at 4°C with primary antibodies overnight as indicated in each figure, and the immunoreactivity was probed with rhodamine red X- or Oregon green 488-conjugated secondary antibodies (see Table S1 in Supplementary Material).

The data were expressed as mean ± SD and analyzed by the one-way analysis of variance procedure followed by least significant difference post hoc tests or Student’s t-tests for three groups, and Student’s t-test for two groups using SPSS 12.0 statistical software (SPSS Inc., Chicago, IL, USA). A p value of <0.05 was considered as statistically significant in all experiments.

The transcription and expression of I2PP2A is significantly increased in the AD brains (Tanimukai et al., 2005), and increasing I2PP2A by AAV transfection in rat brain induced AD-like pathology and cognitive impairment (Wang et al., 2010). We found that I2PP2A protein level was significantly increased in htau transgenic mice compared with the wild-type mice (Figure 1A), while intracranial injection of Lenti - siI2PP2A into the hippocampus of htau transgenic mice, a recognized AD-like animal model for tau pathology (Polydoro et al., 2009), reduced the I2PP2A level to ~45% of the control level at 4 weeks after the injection (Figure 1B). Simultaneously, the phosphorylation level of tau at Thr-205 (pT205), Thr-231 (pT231), Ser-396 (pS396), and Ser-396/404 (PHF-1) epitopes was significantly reduced compared with the htau mice injected with the scrambled Lenti - siI2PP2A controls (Figures 2A–C). By silver staining, we observed that the accumulation of argyrophilic substances was also significantly decreased by expression of Lenti - siI2PP2A (Figure 2D).

A previous study has demonstrated that the htau transgenic mice show learning and memory deficits at 12 months (Polydoro et al., 2009). By step-down avoidance test, we observed that silencing I2PP2A improved both the STM and the LTM (Figure 3). These data demonstrate that silencing I2PP2A by Lenti - siI2PP2A could antagonize tau pathology and memory deficits in htau transgenic mice.

To explore the molecular bases underlying the improved memory by silencing I2PP2A, we analyzed the dendritic morphology and spine density of the neurons in hippocampal CA3 region of the htau transgenic mice using Golgi stain. The dendritic length and the number of branches were assessed as a measure of dendritic complexity. We found that silencing I2PP2A increased the dendritic length, the number of branches, and the density of the dendritic spines (Figure 4), suggesting that silencing I2PP2A promotes dendritogenesis.

Inhibitor-2 of protein phosphatase-2A was originally identified to regulate PP2A (Li et al., 1995, 1996), therefore we first measured PP2A activity after silencing I2PP2A. As expected, silencing I2PP2A significantly increased PP2A activity (Figure 5A). Our recent data show that knockdown of I2PP2A decreases GSK-3β protein level in HEK293 cell lines (Liu et al., 2012), the most implicated tau kinase (Ishiquro et al., 1993; Takashima et al., 1996; Avila and Díıaz-Nido, 2004; Takashima, 2006), and thus we also measured the alteration of GSK-3β in hippocampus of the mice after silencing I2PP2A. We observed that silencing I2PP2A could inhibit GSK-3β, demonstrated by the reduced total level of GSK-3β (tGSK-3β) and elevation of Ser9-phosphorylated GSK-3β (pS9-GSK-3β; Figure 5B). To further explore how GSK-3β activity is regulated, we measured the alteration of Akt and PKA, the known kinases regulating Ser9 phosphorylation of GSK-3β (Cross et al., 1995; Shaw et al., 1997; Fang et al., 2000). We found that silencing I2PP2A activated Akt with increased levels of phosphorylated Akt at Ser473 epitope and total Akt, and PKA with increased levels of PKAα (catalytic subunit) and PKAIIα (regulatory subunit) and decreased PKAIβ (regulatory subunit) level in the hippocampus (Figures 5C,D).

To further verify the role of I2PP2A silencing in the regulation of GSK-3β, we constructed pSUP - siI2PP2A and transfected the plasmid into HEK293/tau or N2a/tau cells with pSUP (empty vector) and pSUP-siC (control vector of siRNA) as controls. As observed in htau transgenic mice, silencing I2PP2A significantly decreased the I2PP2A level in HEK293/tau cells (Figures 6A,B) and N2a/tau cells (Figure 6D). Simultaneously, the activity of PP2A was restored after I2PP2A knockdown in HEK293/tau and N2a/tau cells (Figures 6C,E). I2PP2A knockdown also down-regulated GSK-3β, demonstrated by the reduction of total GSK-3β (tGSK-3β) protein and mRNA levels, and elevation of the pS9-GSK-3β (the inactive form) in HEK293/tau cells (Figures 7A,B) and N2a/tau cells (Figure 7C). These in vitro data further confirm that silencing I2PP2A not only activates PP2A but also inhibits GSK-3β.

Akt and PKA are known kinases to regulate Ser9 phosphorylation of GSK-3β (Cross et al., 1995; Shaw et al., 1997; Fang et al., 2000). Therefore, we studied the role of Akt and PKA in phosphorylating (inhibiting) pS9-GSK-3β. We found unexpectedly that the total Akt level and the phosphorylated Akt at Thr308 and Ser473 (active form) decreased by silencing I2PP2A (Figures 8A,B), suggesting that Akt activity was decreased by I2PP2A knockdown. These data ruled out the role of Akt in phosphorylating GSK-3β during I2PP2A knockdown. However, the catalytic subunit α (PKAα) and the regulatory subunit IIα (PKAIIα) of PKA increased, whereas the regulatory subunit Iβ (PKAIβ) of PKA decreased after silencing I2PP2A in HEK293/tau cells (Figures 8A,B). It is known that PKA is activated when the catalytic subunit is released from the tetrameric holoenzyme, which is modulated by the binding capacity of catalytic subunit to the regulatory subunits. Therefore, we measured the interactions of PKAα with its regulatory subunits by co-immunoprecipitation assay. The results showed that the association level of PKAα with PKAIβ decreased, whereas the binding of PKAα with PKAIIα was not altered after I2PP2A knockdown (Figure 8C). To confirm the alteration of PKA activity, we used ELISA assay, and a significantly increased PKA activity was detected by I2PP2A knockdown (Figure 8D). These data indicate that silencing I2PP2A can also activate PKA.

To further verify the role of PKA in GSK-3β inhibition induced by silencing I2PP2A, we used R_p-cAMPS, a specific inhibitor of PKA, after transfection of siI2PP2A. We found that simultaneous application of R_p-cAMPS abolished the siI2PP2A-induced inhibitory phosphorylation of GSK-3β at Ser9 (Figure 8E).

We also studied whether silencing siI2PP2A could attenuate tau hyperphosphorylation induced by OA (PP2A inhibitor) in vitro. We transfected pSUP - siI2PP2A into HEK293/tau or N2a/tau cells, and treated the cells with OA (25 nM) for 24 h. Then, we detected tau phosphorylation by Western blotting. We found that silencing I2PP2A significantly reduced tau phosphorylation at Ser-199 (pS199), Thr-205 (pT205), Ser-214 (pS214), Ser231 (pS231), Ser-396 (pS396), and Ser-404 (pS404), and increased the level of the unphosphorylated tau at Ser-198/202 (tau-1) induced by OA in HEK293/tau (Figures 9A,B). The immunofluorescence data confirmed the same results (Figure 9C). We also found that knockdown I2PP2A significantly decreased tau phosphorylation induced by OA in N2a/tau cells (Figure 9D).