Selective Degeneration of Entorhinal-CA1 Synapses in Alzheimer's Disease via Activation of DAPK1

Excitatory pyramidal neurons in the entorhinal cortical layer II region (ECIIPN) form functional excitatory synapses with CA1 parvalbumin inhibitory neurons (CA1PV) and undergo selective degeneration in the early stages of Alzheimer's disease (AD). Here, we show that death-associated protein kinase 1 (DAPK1) is selectively activated in ECIIPN of AD mice. Inhibition of DAPK1 by deleting a catalytic domain or a death domain of DAPK1 rescues the ECIIPN-CA1PV synaptic loss and improves spatial learning and memory in AD mice. This study demonstrates that activation of DAPK1 in ECIIPN contributes to a memory loss in AD and hence warrants a promising target for the treatment of AD. SIGNIFICANCE STATEMENT Our recent study reported that excitatory pyramidal neurons in the entorhinal cortical layer II region (ECIIPN) target to CA1 parvalbumin-type inhibitory neurons (CA1PV) at a direct pathway and are one of the most vulnerable brain cells that are selectively degenerated in the early stage of Alzheimer's disease (AD). Our present study shows that death-associated protein kinase 1 (DAPK1) is selectively activated in ECIIPN of AD mice. Inhibition of DAPK1 by deleting a catalytic domain or a death domain of DAPK1 rescues the ECIIPN-CA1PV synaptic loss and improves spatial learning and memory in the early stage of AD. These data not only demonstrate a crucial molecular event for synaptic degeneration but also provide a therapeutic target for the treatment of AD.


Introduction
Alzheimer's disease (AD) is the most common form of dementia in the elderly, affecting Ͼ10 million people in China and Ͼ35 million people worldwide. The deposition of senile plaques that primarily consist of amyloid-␤ (A␤) peptide is a major pathological hallmark in the brains of AD patients and has long been considered to be associated with a progressive loss of central neurons in certain regions of the brain (Tanzi et al., 1987;Goate et al., 1991;Jack et al., 2010;Duffy et al., 2015). However, recent studies have shown that impairments in learning and memory, the early clinical signs of AD, are caused by synaptic dysfunction rather than neuronal cell loss. For example, in AD patients, cognitive decline is closely associated with a reduction in the number of presynaptic glutamatergic terminals (Gomez-Isla et al., 1996;Kamenetz et al., 2003;Oddo et al., 2003). Tg2576-APPswe mice (AD mice), which carry a transgene encoding the 695 amino acid isoform of the human A␤ precursor protein with the Swedish mutation and exhibit plaque pathologies similar to those in AD patients, show decays in synaptic transmission and impairments in spatial learning and memory in an age-dependent manner Jacobsen et al., 2006;Scheff et al., 2007).
However, which of the many synapses in the brain undergo selective degeneration during the early stages of AD and whether this selective degeneration contributes directly to the loss of spatial learning and memory are still unknown.
Excitatory pyramidal neurons in the entorhinal cortex (EC, EC PN ), which primarily target to the hippocampus, are the most vulnerable brain cells in the early stages of AD (Hsia et al., 1999;Yassa, 2014;Yang et al., 2016). The EC PN are largely distributed in the EC layer II (ECII PN ) and III (ECIII PN ) regions. These neurons innervate excitatory pyramidal neurons and parvalbumin (PV) inhibitory GABA-containing neurons in the CA1 hippocampus (CA1 PV ) and are associated with spatial and temporal associative memories (Kitamura et al., 2014;Yang et al., 2016). Our recent study showed that amyloid deposition occurs in the brain of AD mice at 8 months of age, whereas synaptic transmission between ECII PN and CA1 PV undergoes degeneration at 6 months of age (Yang et al., 2016). This finding indicates that synaptic degeneration is not associated with the presence of amyloid plagues. Yet, the molecular mechanisms underlying degeneration of ECII PN synapses remain unknown. Our present studies demonstrate that death-associated protein kinase 1 (DAPK1) became activated selectively in the ECII PN of AD mice. We show that activation of DAPK1 is responsible for a selective degeneration of ECII PN -CA1 PV synapses and that inhibition of DAPK1 is therapeutically effective for the intervention of spatial learning and memory declines in AD.

Materials and Methods
Animals. All mice used in this study were bred and reared under the same conditions in the University's core animal facility in accordance with institutional guidelines and the Animal Care and Use Committee (Huazhong University of Science and Technology, Wuhan, China). The mice were housed in groups of 3-5 per cage and maintained with a 12 h light-dark cycle, with lights on at 8:00 A.M., at consistent ambient temperature (22 Ϯ 1°C) and humidity (50 Ϯ 5%). Tg2576-APPswe mice (the AD mice), which expressed a mutant form of A␤ precursor protein (APP) (isoform 695) with the Swedish mutation (KM670/671NL), were purchased from the The Jackson Laboratory. In the present study, all the AD mice were identified as homozygous. Male mice were used in this study.
Generation of the mutant mice. To generate the AD/ECII PN ChR2ϩ mice, we expressed ChR2-eGFP in the ECII PN of mutant mice that had a loxPflanked STOP sequence followed by ChR2 (E123A)-eGFP. The Rosa-CAG-Flag-ChR2eGFP-WPRE targeting vector was designed with a CMV-IE enhancer/chicken ␤-actin/rabbit ␤-globin hybrid promoter (CAG), an FRT site, a loxP-flanked STOP cassette, a Flag-eGFP sequence, a woodchuck hepatitis virus post-transcriptional regulatory element (WPRE; to enhance the mRNA transcript stability), a poly-A signal, and an attB/att-flanked PGK-FRT-Neo-poly-A cassette. This entire construct was inserted into the Gt(ROSA)26Sor locus via electroporation into C57BL/6-derived embryonic stem (ES) cells. The targeted ES cells were selected and injected into C57BL/6 blastocysts, and chimeric animals were then bred to C57BL/6 mice (ChR2 loxP /loxP ).
Next, we created the rAAV1/2-D28K-Cre virus particles. We designed the rAVE-D28K-Cre vector by inserting a Cre recombinase sequence immediately downstream of the calbindin 1 (or calbindin-D28K, D28K ) translational STOP codon via Apal/Kpnl. The rAVE plasmids were cotransfected with AAV helper1/2 mixers into HEK293 cells to generate rAAV1/2-D28K-Cre virus particles with a high titer (Ͼ5 ϫ 10 12 genomic particles/ml), as described previously (Tu et al., 2010;Pei et al., 2015;Yang et al., 2016). The virus particles (1.5 l) were bilaterally injected into the EC superficial layer (anteroposterior 4.8 mm, mediolateral 2.8 mm, dorsoventral 3.5-4.0 mm) in the AD/ChR2 loxP /loxP and control/ChR2 loxP /loxP mice, thus leading to ChR2-eGFP expression in the ECII PN of the AD or control mice (AD/ECII PN ChR2ϩ and control/ECII PN ChR2ϩ mice). To generate the AD/ECII PN CDϪ mice, we first created a mutant strain of mice with selective deletion of the catalytic domain (CD) of DAPK1 (DAPK1 CDloxP/loxP mice) by targeting the exon 2 region, as described previously (Pei et al., 2015). The DAPK1 CDloxP/loxP mice were crossed with the AD/ECII PN ChR2ϩ mice, thus resulting in AD/ECII PN CDϪ mice, in which the DAPK1 CD was selectively deleted in the ChR2 ϩ -ECII PN . To generate the AD/ECII PN DDϪ mice, we created a mutant strain of mice with a selective deletion of the death domain (DD) of DAPK1 (DAPK1 DDloxP/loxP mice). In brief, an FRT-flanked Neo resistance positive selection cassette was inserted downstream of exon 27; one loxP site was introduced upstream of exon 26 and another loxP site was introduced downstream of the Neo cassette. After linearization, the targeting vector was transfected into C57BL/6J embryonic stem cells via electroporation. Six positive clones were identified by Southern blotting with a 5Ј probe, a 3Ј probe, and a Neo probe. Two positive clones were injected into BALB/c blastocysts and implanted into pseudopregnant females. Chimeric mice were crossed with C57BL/6J mice to obtain F1 mice carrying the recombined allele containing the floxed DAPK1 allele and the Neo selection cassette. These mice were mated with Flp recombinaseexpressing C57BL/6J Flp mice to remove the Neo resistance cassette and to generate a line of Neo-excised DAPK1 DDloxP/loxP mice. When these mice were bred with the AD/ECII PN ChR2ϩ mice, the offspring were AD/ ECII PN DDϪ mice, in which the DAPK1 DD was selectively deleted in the ChR2 ϩ -ECII PN .
Single-cell Western blots, kinase assays, and coimmunoprecipitation. EC tissues were isolated from the AD/ECII PN ChR2ϩ mice, the AD/ ECII PN CDϪ mice, the AD/ECII PN DDϪ mice and their respective controls and sliced and digested in buffer containing 10 mM Tris-Cl, pH 7.6, 50 mM NaF, 1 mM Na 3 VO 4 , 1 mM edetic acid, 1 mM benzamidine, 1 mM PMSF, 1 mg/10 ml papain, and a mixture of aprotinin, leupeptin, and pepstatin A (10 g/ml each) for 30 min. The cell suspension of ECII PN eGFPϩ was automatically isolated using an S3e Cell Sorter (Bio-Rad). The isolated ECII PN were homogenized and diluted with a buffer containing 200 mM Tris-Cl, pH 7.6, 8% SDS, and 40% glycerol. The protein concentration was determined using a BCA kit (Pierce). Final concentrations of 10% ␤-mercaptoethanol and 0.05% bromophenol blue were added, and the samples were boiled for 10 min in a water bath. The proteins in the extracts were separated by 10% SDS-PAGE and transferred to nitrocellulose membranes. The blots were scanned by using an Infrared Imaging System (Odyssey, LI-COR). The blots were incubated with antibodies against DAPK1 (1:500, Epitomics, catalog #3798-1). The band densities were quantitatively analyzed by using Kodak Digital Science 1D software (Eastman Kodak).
Immunocomplex kinase assay was used to determine the catalytic activity of DAPK1 in the lysates and myosin light chain (MLC) that was phosphorylated at Ser-18/Thr-20 ( pMLC) by activated DAPK1 was blotted with anti-pMLC (1:1000, Millipore). The enzymatic activity of DAPK1 was expressed as the ratio of pMLC to DAPK1 protein under the same conditions. Immunoprecipitation was used to determine the association of DAPK1 with its substrates. The lysates (ϳ200 g protein) were incubated with nonspecific IgG (2 g) or polyclonal rabbit anti-DAPK1 antibody (2 g, Millipore) overnight at 4°C, and this was followed by the addition of 40 l of Protein G-Sepharose (Sigma) and incubation for 3 h at 4°C. The precipitates were washed four times with lysis buffer, denatured with SDS sample buffer, and separated by 12% SDS-PAGE. The proteins were transferred onto nitrocellulose membranes using a Bio-Rad mini-protein-III wet transfer unit overnight at 4°C. The transferred membranes were then incubated with blocking solution [5% nonfat dried milk dissolved in TBST buffer containing 10 mM Tris-HCl, 150 mM NaCl, and 0.1% Tween 20 for 1 h at room temperature, washed three times, and incubated with anti-goat primary antibody against Tau (1: 1000, Santa Cruz Biotechnology), anti-rabbit antibody against Erk (1: 1000; Millipore), or anti-DAPK1 (1:1000, Sigma) for 1 h at room temperature. The membranes were washed three times with TBST buffer, incubated with the appropriate secondary antibodies (1:1000 dilution) for 1 h, and washed four times. Signal detection was performed with an enhanced chemiluminescence kit (GE Healthcare). The lanes marked "input" were loaded with 10% of the starting material used for immunoprecipitation.
Extracellular single-unit recording. Extracellular single-unit and local field power spectrum recordings were made from the CA1 neurons (anteroposterior Ϫ1.7 mm, mediolateral 1.0 mm, and dorsoventral 0.5-0.8 mm target to CA1). Mice were connected to the recording equipment via AC-coupled unity-gain operational amplifiers (Plexon). Signals were amplified 4000-to 8000-fold. The spikes and local field potentials were recorded simultaneously and isolated by using a 250 Hz low-pass filter and a 250 Hz high-pass filter and the commercial software OmniPlex (Plexon). Spike sorting was performed off-line using graphical clustersorting software (Offline Sorter, Plexon). To estimate the quality of the cluster separation, we calculated the isolation distance and the L-ratio using Plexon SDK. Only the units with an L-ratio Ͻ0.05 and a distance Ͼ15 were included.
To isolate and analyze spike units from individual CA1 PN versus the CA1 PV , we calculated the valley-to-peak time and the half-width of the spikes. Spikes from CA1 PN were identified and distinguished from those of CA1 IN according to the duration of the negative spike, the firing pattern (complex spikes), and the low average firing rate. Spikes from CA1 PV were validated as described previously (Yang et al., 2016). The average firing rate was expressed as the total number of spikes divided by the total length of the recording period.
Morris Water Maze. A pool 1.5 m in diameter was filled with water that was made opaque with white nontoxic ink and maintained at 25.0°C. Animals were brought to the behavior room (where they were housed for the duration of the training), handled for 1-2 d, and trained, as described previously (Tu et al., 2010;Pei et al., 2015;Yang et al., 2016). The entire protocol lasted 7 d. The first training day consisted of a probe trial followed by a "visible platform" trial, in which the platform was indicated by a red flag. Next, mice were subjected to their first "hidden platform" learning trial, during which they were allowed to rest on the platform for 30 s before being released from one of the pool's starting points (north, south, east, or west). The animals were allowed 60 s to find the platform and were allowed to stay there for 30 s; if an animal did not find the platform within 60 s, it was removed from the water and placed on the platform for 30 s. During days 1-6, four trials were performed corresponding to the four different randomized release points. Therefore, the total training took 6 d followed by a final 24 h probe on the seventh day.
Open-field tests and rotarod tests. Locomotor activity was assessed in clear boxes measuring 100 cm ϫ 100 cm that were outfitted with photobeam detectors for monitoring horizontal and vertical activity, as described previously (Tu et al., 2010;Pei et al., 2015;Yang et al., 2016). The data were collected using a PC and were analyzed using the MED Associates Activity Monitor Data Analysis software. Mice were placed in a corner of the open-field apparatus and allowed to move freely. The variables recorded included resting time (seconds), ambulatory time (sec-onds), vertical/rearing time (seconds), jump time (seconds), stereotypic time (seconds), and average velocity (centimeters per second). The mice were not exposed to the chamber before testing. The data were recorded for each animal during 30 min intervals. For the rotarod test, mice were subjected to a 1 week learning period, after which they were able to perform on an accelerating rotarod. The test was then performed twice per week until the mice were unable to remain on the rotating bar for Ͼ10 s in three consecutive attempts, which was defined as rotarod failure.
Data analysis. All variance values in the text and figure legends are represented as the mean Ϯ SEM. Parametric tests, including t tests and two-way ANOVAs, were used when assumptions of normality and equal variance (F test) were met; two-way repeated-measures ANOVA was used to analyze the Morris Water Maze data. Differences were considered statistically significant when p Ͻ 0.05.

Activation of DAPK1 in ECII PN of AD mice
To determine the molecular mechanisms underlying the selective degeneration of ECII PN synapses in AD mice, we focused our studies on DAPK1, an enzyme critical for synaptic losses in central neurons (Tu et al., 2010). To functionally and morphologically identify individual ECII PN , we generated AD/ECII PN ChR2ϩ mice, in which channelrhodopsin-2-E123A (ChR2), a modified version of a light-gated cation channel, and an eGFP were selectively expressed in the vulnerable ECII PN of the AD mice. A mouse model of AD was chosen because these mice carry a transgene coding for the 695 amino acid isoform of the human Alzheimer A␤ precursor protein with the Swedish mutation and exhibit plaque pathologies similar to those in AD patients.
Specifically, we created a conditional line of mutant mice that expressed a double-floxed inverted open reading frame of ChR2-eGFP (ChR2 loxP/loxP mice; Fig. 1A). We also constructed a Type 1/2 recombinant adeno-associated (rAAV1/2)-D28K-Cre virus vector, in which Cre recombinase was expressed under the control of the D28K promoter (Fig. 1B). Virus particles at high titer were injected directly into the ECII of the AD/ChR2 loxP/loxP mice, resulting in the ChR2-eGFP expression specifically in the ECII PN (Fig. 1B). ECII PN cells were named as island cells, in which Wfs1 is expressed (Kitamura et al., 2014). Consistent with this previous study, we showed that ECII PN that directly innervate CA1 PV were labeled with an antibody against Wfs1 protein (Fig.  1C). Thus, ECII PN ChR2ϩ neurons comprise a group of excitatory island cells.
We next isolated ECII PN ChR2ϩ cells from the brain tissues of the AD/ECII PN ChR2ϩ mice using flow cytometry and cell sorting techniques. To determine the catalytic activity of DAPK1 in the purified ECII PN ChR2ϩ cells, we applied an immunocomplex kinase assay with MLC as an endogenous substrate of DAPK1 (Tu et al., 2010). We precipitated DAPK1 protein complex in the cell lysates of the ECII PN ChR2ϩ cells from the AD/ECII PN ChR2ϩ mice and the controls (the control/ECII PN ChR2ϩ mice) using an antibody against DAPK1. The precipitates were then blotted with antibodies against a pMLC and DAPK1, respectively (Fig. 1D). Our data revealed that the levels of pMLC were increased in the ECII PN ChR2ϩ cells from the AD/ECII PN ChR2ϩ mice after 150 d of age (Fig. 1D), compared with the age-matched control/ ECII PN ChR2ϩ mice (Fig. 1E). The pMLC was catalyzed by an activated DAPK1 as MLC kinase (MLCK) that also targets MLC at serine-20 was undetectable in the precipitates from the ECII PN ChR2ϩ cells of the AD/ECII PN ChR2ϩ mice (Fig. 1F ). An increase of the pMLC was found in the ECII PN ChR2ϩ cells only, but not in the frontal cortical neurons from the AD/ECII PN ChR2ϩ mice (Fig. 1G). Together, these data demonstrate that DAPK1 is activated selectively in ECII PN cells of AD mice during aging.

Genetic inhibition of DAPK1 in ECII PN of AD mice
We next determined whether activation of DAPK1 in the ECII PN of AD mice contributes to impairments in synaptic transmission along the ECII PN -CA1 PV pathway. We developed two independent approaches to inhibiting DAPK1 function in the ECII PN of AD mice. First, we generated a mutant strain of mice (DAPK1 CDloxP/loxP ) in which a double-floxed inverted open reading frame of DAPK1 with a CD deletion was expressed ( Fig. 2A,B). When these mice were bred with AD/ECII PN ChR2ϩ mice, which Cre recombinase was expressed in the ECII PN ChR2ϩ cells (for details, see Fig. 1B), the CD of DAPK1 (DAPK1 CDϪ ) was selectively deleted in the ECII PN ChR2ϩ of the offspring (AD/ECII PN CDϪ mice; Fig. 2C-E). To verify the suc-cessful generation of AD/ECII PN CDϪ mice, we performed two independent lines of the studies. First, we stained the sections from the AD/ECII PN CDϪ mice with antibody against D28K protein and showed that the DAPK1 CDϪ mutant protein in the ECII PN ChR2ϩ cells was recognized by anti-D28K (Fig. 2D). Second, we isolated the ECII PN ChR2ϩ cells from the AD/ECII PN CDϪ mice using flow cytometry and cell sorting techniques. The cell lysates were prepared from the isolated ECII PN ChR2ϩ cells and precipitated with anti-DAPK1. The precipitates were blotted with antibodies against pMLC and DAPK1 proteins, as indicated (Fig. 2E). Our data revealed a complete loss of DAPK1 enzymatic activity in the ECII PN ChR2ϩ cells of the AD/ECII PN CDϪ mice (Fig. 2E). Next, we generated the mutant mice (DAPK1 DDloxP/loxP mice) in which a double-floxed inverted open reading frame of DAPK1 lacking a DD was expressed (Fig. 2 F, G). When these mice were ChR2ϩ cells were isolated from the brain sections of the AD/ECII PN ChR2ϩ mice (D) and the control/ECII PN ChR2ϩ mice (E), respectively, when they were at 120, 150, 180, and 210 d of age. The cell lysates were then prepared from the isolated ECII PN ChR2ϩ cells and precipitated with anti-DAPK1. The precipitates were blotted with antibodies against pMLC or DAPK1, as indicated. In bar graphs, the band intensity was normalized to that of the DAPK1 protein from a mouse at 120 d of age (defined as 1.0). Data are mean Ϯ SEM (n ϭ 5). *p Ͻ 0.001 (t tests). F, G, the pMLC is catalyzed by an activated DAPK1 but not MLCK in the ECII PN ChR2ϩ cells from AD mice. Blots of the precipitates from the ECII PN ChR2ϩ (F ) and the frontal cortical cells (G) extracts of the AD/ECII PN ChR2ϩ mice at 120, 150, 180, and 210 d of age with antibodies against MLCK, pMLC, or DAPK1, as indicated. Similar results were seen in each of the four experiments.
crossed with the AD/ECII PN ChR2ϩ mice, the DD of DAPK1 was selectively deleted (DAPK1 DDϪ ) in the ECII PN ChR2ϩ of the offspring (AD/ECII PN DDϪ mice). The successful expression of DAPK1 DDϪ mutant protein was confirmed by blotting the cell lysates from the ECII PN ChR2ϩ cells of the AD/ECII PN DDϪ mice with antibody against N-terminal fragment of DAPK1 (N-DAPK1; Fig.  2H). We next examined the enzymatic activity of DAPK1 DDϪ mu-tant protein. We isolated the ECII PN ChR2ϩ cells from the AD/ ECII PN DDϪ mice using flow cytometry and cell sorting techniques. The cell lysates were prepared from the isolated ECII PN ChR2ϩ cells and precipitated with anti-DAPK1. The precipitates were blotted with antibodies against pMLC, tubulin, Tau, and ERK proteins, as indicated (Fig. 2I,J). Our data revealed that the expression of DAPK1 DDϪ did not alter the catalytic activity of DAPK1 (Fig. 2I), The Tg3 mice were generated by crossing the ChR2 loxP/loxP mice with the AD mice and the DAPK1 CDloxP/loxP mice and were stereotaxically injected with the rAAV1/2-D28K-Cre virus particles into the ECII region, resulting in the AD/ECII PN CDϪ mice. D, Representative images showing the ECII PN from the AD/ECII PN CDϪ mice that expressed the DAPK1 CDϪ mutant protein (green, ChR2) and that were stained with an antibody against the D28K protein (pink). Similar results were seen in each of the five experiments. E, Blots of the ECII PN extracts from the AD/ECII PN CDϪ mice using antibodies against pMLC or DAPK1, as indicated. The band intensity was normalized to that of the DAPK1 protein from mice at 90 d of age (defined as 1.0). Data are mean Ϯ SEM (n ϭ 5). F, Constructs for the generation of the DAPK1 DDloxP/loxP mice. G, H, Representative images showing the expression of mutant transcripts (G) and protein (H ) from the DAPK1 DDloxP/loxP mice. H, An antibody against the N-terminal region of DAPK1 (N-DAPK1) recognizes the wild-type (ϳ160 kDa) and mutant proteins (ϳ140 kDa). Similar results were seen in each of the five experiments. I, Deletion of the DAPK1 DD does not affect the catalytic activity of DAPK1. Blots of the ECII PN extracts from the AD/ECII PN DDϪ mice using antibodies against pMLC or N-DAPK1, as indicated. The band intensity was normalized to that of the N-DAPK1 protein from mice at 90 d of age (defined as 1.0). Data are mean Ϯ SEM (n ϭ 5). J, Deletion of the DAPK1 DD inhibits the association of DAPK1 protein with its substrates. The ECII PN cell lysates from the AD/ECII PN DDϩ (DD ϩ ) or the AD/ECII PN DDϪ (DD Ϫ ) mice were precipitated with nonspecific IgG or antibodies against DAPK1. The precipitates were then blotted with antibodies against N-DAPK1, Tau protein, or ERK protein, as indicated. Input: 10 g of protein from the ECII PN cell lysates without precipitation was loaded. Similar results were seen in each of the four experiments.

mV intervals) of stimulation of ECII PN
ChR2ϩ . Data are mean Ϯ SEM (n ϭ 25 recordings/5 mice/group). *p Ͻ 0.001 (t tests). Representative recordings above the graph are the average of 12 sweeps of the evoked EPSCs at the low (a) and high (b) stimulus intensities. B, Representative images show the ChR2-eGFP-labeled terminals (green) of the ECII PN in the CA1 hippocampus stained with anti-PV antibody (pink). C, Bar graphs represent the numbers (terminals/0.2 cm 2 ) of the ChR2-eGFP-labeled terminals in the stratum lacunosum-moleculare region and the PV-labeled spines (spines/10 m dendritic branches) in the stratum lacunosum-moleculare region. Data are mean Ϯ SEM (n ϭ 5 mice/group). *p Ͻ 0.001 (t tests). D-F, A working model showing that inhibition of DAPK1 restores excitatory and inhibitory balance in the CA1 circuits of AD mice. Under the physiological conditions (D), ECII PN form direct excitatory synapses with CA1 PV and balance the excitatory/ inhibitory synaptic transmission in CA1 circuits. During the disease progression of AD (E), ECII PN -CA1 PV synapses are degenerated. This degeneration disables the excitatory and inhibitory balance as a consequence of a loss of inhibitory inputs from CA1 PV to CA1 PN . Inhibition of DAPK1 in the ECII PN (F ) effectively intervenes in the degeneration of EC PN -CA1 PV synapses and restores the excitatory and inhibitory synaptic balance in AD mice. G, H, Genetic inhibition of DAPK1 restores the balance between excitation and inhibition in CA1 PV (G) and CA1 PN (H ) cells of AD mice at 180 d of age. Representative recordings of spike units of CA1 neurons in freely moving mice. The spikes in CA1 PV (G) versus CA1 PN (H ) were isolated based on the valley-to-peak time and the half-width of the spikes. The averaged frequencies of action potential firings of CA1 PV (G) and CA1 PN (H ) in freely moving mice at 180 d of age were summarized in bar graphs. Data are mean Ϯ SEM (n ϭ 11 mice per group). *p Ͻ 0.001 (t tests). but this deletion blocked the binding of DAPK1 to DAPK1 substrates, such as the Tau and ERK proteins (Fig. 2J).

Protection against ECII PN synaptic decay in AD mice
To determine whether the genetic inhibition of DAPK1 might avert the ECII PN synaptic decay in AD mice, we performed whole-cell patch-clamp recordings with the CA1 PV (Fig. 3A) from the AD/ECII PN CDϪ and AD/ECII PN DDϪ mice at 180 Ϯ 5 d of age as well as age-matched controls (AD/ECII PN ChR2ϩ , the control/ECII PN DDϪ , and the control/ECII PN ChR2ϩ mice). EPSCs were evoked by delivering blue laser light directly to the ECII PN ChR2ϩ in brain slices. To exclude disturbances from other projections (i.e., the ECII PN -DG-CA3-CA1 PV indirect pathway), we cut off the connections of the dentate gyrus to the CA3 cells in the slices (Fig. 3A) and recorded NMDA-mediated EPSCs at a holding potential of 60 mV. The recordings were performed in the pres-  , E), 120 (F ), and 180 (G) days old of age. D, Representative hot spot of path tracings taken during the probe trials on day 8. Data are mean Ϯ SEM (n ϭ 11 mice per group, F (5,60) ϭ 3.97). *p Ͻ 0.001. H, Deletion of DAPK1 DD in the ECII PN of AD mice improves spatial information acquisition. The latency and swim length to reach a hidden platform are plotted against the blocks of trials. Data are mean Ϯ SEM (n ϭ 11 mice per group, F (5,60) ϭ 4.13). *p Ͻ 0.001. I, Deletion of DAPK1 DD in the ECII PN of AD mice improves spatial memory. The percentage of time spent in search of a hidden platform in each quadrant during the probe trial. Data are mean Ϯ SEM (n ϭ 11 mice per group, F (5,60) ϭ 3.11). *p Ͻ 0.001. ence of 20 M CNQX and 20 M bicuculline to block both AMPA receptors and Type A GABA A receptor-mediated responses. Under this circumstance, we were able to record monosynaptic transmission between ECII PN and CA1 PV cells in the slices. We then analyzed the mean amplitudes of the evoked EPSCs with the increases of the stimulus intensities (Fig. 3A). We demonstrated that the genetic inhibition of DAPK1 specifically in the ECII PN did not alter synaptic transmission in the control mice (the peak values of the mean amplitudes were 99 Ϯ 13 pA in the control/ECII PN ChR2ϩ mice vs 92 Ϯ 15 pA in the control/ ECII PN CDϪ mice), but it was effective in reversing the ECII PN -CA1 PV synaptic decline in the AD mice (peak values of the mean amplitudes were 65 Ϯ 8 pA in the AD/ECII PN ChR2ϩ mice vs 93 Ϯ 9 pA in the AD/ECII PN CDϪ mice). Next, we examined the structure of ECII PN -CA1 PV synapses in the AD/ECII PN CDϪ mice (Fig. 3B). At presynaptic sites, the excitatory terminals in the stratum lacunosum-moleculare, an afferent axon terminal zone of the ECII PN ChR2ϩ in the CA1 hippocampus, were analyzed (Fig. 3B). The density of ChR2-eGFPlabeled terminals (50 m ϫ 200 m) in the AD mice at 180 Ϯ 5 d of age was 29.1 Ϯ 3.2% lower than that of the age-matched controls (Fig. 3C). This reduction in the number of excitatory synaptic terminals in the the stratum lacunosum-moleculare region of the AD mice was completely rescued in both the AD/ ECII PN CDϪ and AD/ECII PN DDϪ mice (Fig. 3C). At postsynaptic sites, we stained the sections with an antibody against the PV protein (Fig. 3B), which revealed that the dendritic branch spines (50 m segment) from the antibody-labeled CA1 PV cells (50 seg-ments per animal) in the stratum lacunosum-moleculare region of the AD mice decreased by 35.1 Ϯ 3.9% compared with those in the age-matched controls (5.3 Ϯ 0.63 vs 3.3 Ϯ 0.36, mean Ϯ SEM, n ϭ 50 dendritic branches/5 mice/group, p Ͻ 0.001; Fig. 3C). The inhibition of DAPK1 completely rescued the synaptic loss of the CA1 PV in the AD mice (5.1 Ϯ 0.59 in the AD/ECII PN CDϪ mice, 4.9 Ϯ 0.62 in the AD/ECII PN DDϪ mice vs 5.2 Ϯ 0.58 in the controls, mean Ϯ SEM, n ϭ 50 dendritic branches/5 mice/group, p Ͼ 0.05; Fig. 3C).
Impairments of synaptic transmission along the ECII PN -CA1 PV excitatory pathway, which primarily target the dendrites of CA1 excitatory pyramidal neurons (CA1 PN ) disrupt the excitatory and inhibitory balance in the CA1 neural circuits of AD mice (Lesne et al., 2006;Mucke, 2007;Palop and Mucke, 2009). Therefore, protecting the ECII PN -CA1 PV pathway from synaptic loss in AD mice should restore the excitatory and inhibitory balance in the CA1 circuits ( Fig. 3D-F ). To test this hypothesis, we monitored the activity of CA1 neuronal cells in freely moving mice at 180 Ϯ 5 d of age by using extracellular single-unit recording techniques (Yang et al., 2016). Action potentials that originated from the CA1 PN versus the CA1 PV of freely moving mice were classified on the basis of the properties of the action potential, as recently described (Yang et al., 2016). In the AD mice, the probability of action potential firing was dramatically reduced in the CA1 PV (11.8 Ϯ 1.6 vs 16.9 Ϯ 1.9, mean Ϯ SEM, n ϭ 36 units/9 mice/group, p Ͻ 0.001; Fig. 3G) and increased in the CA1 PN (5.6 Ϯ 0.68 vs 3.3 Ϯ 0.51, mean Ϯ SEM, n ϭ 36 units/9 mice/ group, p Ͻ 0.001; Fig. 3H ), compared with the age-matched Figure 5. AD mice are normal in learning-unrelated behavioral tests. A, Bar graphs represent that distance moved, rearing counts, and grooming time in the open field tests are identical among groups (mean Ϯ SEM, n ϭ 9 mice per group). B, Bar graphs represent that the performance for foot stay and fault on rotarod tests is comparable between genotypes (mean Ϯ SEM, n ϭ 8 mice per group). C, Mice are normal in elevated plus maze tests. Bar graphs represent that time in open arms is identical between groups (mean Ϯ SEM, n ϭ 9 mice per group). D, Mice are normal in the forced swimming tests. The percentage of time immobile per minute over the whole 5 min trial is identical between groups (mean Ϯ SEM, n ϭ 9 mice per group).
control mice, thus showing that the excitatory and inhibitory balance was disrupted in the CA1 circuits of the AD mice. In the AD/ECII PN CDϪ mice, in which DAPK1 was inactivated specifically in the ECII PN , the excitatory and inhibitory balance in the CA1 circuits was restored; the probabilities of action potential firings in both the CA1 PV (16.2 Ϯ 1.8 vs 17.5 Ϯ 1.8, mean Ϯ SEM, n ϭ 36 units/9 mice/group, p Ͼ 0.05) and the CA1 PN (3.7 Ϯ 0.52 vs 3.7 Ϯ 0.58, mean Ϯ SEM, n ϭ 36 units/9 mice/group, p Ͼ 0.05; Fig. 3 H, G) were identical to those in the age-matched controls. In the AD/ECII PN DDϪ mice, action potential firing in both the CA1 PV and CA1 PN was comparable with that in the age-matched controls. Together, these data demonstrate that the inhibition of DAPK1 by selective deletion of either the DAPK1 CD or DAPK1 DD within the ECII PN prevents the decay of the ECII PN -CA1 PV synaptic transmission and hence restores the excitatory and inhibitory balance in the CA1 circuits of the AD mice.

Improvements in spatial learning and memory in AD mice
We next determined whether preventing the synaptic impairments via the inhibition of DAPK1 in the ECII PN improves spatial learning and memory in the AD mice. We analyzed the task performance of mice at 180 Ϯ 5 d of age in a hidden version of the Morris Water Maze test. The AD/ECII PN CDϪ mice showed better performance in all the measured indices ( Fig. 4A-G); the latency and swim length to reach a hidden platform during the training session ( Fig. 4A-G), and the percentage of time spent in search of a hidden platform in each quadrant during the probe trial ( Fig.  4D-G) were comparable between the AD/ECII PN CDϪ mice (latency ϭ 16.9 Ϯ 1.7 s; length ϭ 192 Ϯ 21 cm; percentage time ϭ 37 Ϯ 3.5) and the control AD/ECII PN CDϩ mice (latency ϭ 15.7 Ϯ 1.3 s; length ϭ 183 Ϯ 19 cm; percentage time ϭ 35.5 Ϯ 3.6). These data indicate that the inhibition of DAPK1 effectively averts the decay of spatial learning and memory in the AD mice. Consistently with this conclusion, the AD/ECII PN DDϪ mice with a deletion of the DAPK1 DD in the ECII PN also exhibited significant improvements in acquiring spatial information during the Morris Water Maze tests (latency ϭ 15.0 Ϯ 1.6 s; length ϭ 191 Ϯ 18 cm; percentage time ϭ 38 Ϯ 3.2 in the AD/ECII PN DDϪ mice vs latency ϭ 17.1 Ϯ 1.6 s; length ϭ 189 Ϯ 20 cm; percentage time ϭ 34.9 Ϯ 3.3 in the AD/ECII PN ChR2ϩ mice; Fig. 4 H, I ). To determine the specific effects of DAPK1 inhibition on spatial learning and memory, we also performed learning-unrelated behavioral tests, including open field, rotarod, elevated plus maze, and forced swimming. Our data revealed that all the groups of mice, including the AD mice at 180 Ϯ 5 d of age, performed normally in all of the learning-unrelated tests, compared with the agematched controls (Fig. 5A-D).

Discussion
ECII PN are some of the earliest affected brain cells in AD, and the selective expression of human mutant APP in the ECII PN of mice impairs spatial learning and memory (Harris et al., 2010). Consistently, our data revealed that DAPK1 was selectively activated in the ECII PN of AD mice and that the genetic inhibition of DAPK1 effectively protected against impairments in the ECII PN -CA1 PV synaptic transmission and improved spatial learning and memory.
The key findings in the present study include the following: (1) DAPK1 is selectively activated in ECII PN in the early stages of AD mice; and (2) the specific inhibition of DAPK1 in the ECII PN of AD mice results in therapeutic effects against declines in spatial learning and memory. DAPK1 is a Ca 2ϩ / calmodulin-dependent protein kinase and was originally identified by functional cloning on the basis of its involvement in interferon-␥-induced apoptosis (Shohat et al., 2001). Previously, we have reported that DAPK1 is activated in central neurons and contributes to neuronal death (Tu et al., 2010). More recently, we have reported that activated DAPK1 directly binds and phosphorylates the Tau protein on Ser262 and induces synaptic degeneration (Pei et al., 2015). Hyperphosphorylation of Tau (tau inclusions, pTau) results in the self-assembly of tangles of paired helical filaments and straight filaments in the brain, which are involved in the early pathogenesis of AD (Alonso et al., 2001;Clavaguera et al., 2009;Hoover et al., 2010;de Calignon et al., 2012;Spires-Jones and Hyman, 2014). In the present study, we found that genetic deletion of the DAPK1 DD in the AD/ECII PN DDϪ mice disrupted binding of DAPK1 to the Tau protein in ECII PN . Thus, the DAPK1-Tau interaction may be a crucial signaling event underlying the early degeneration of excitatory synaptic transmission at ECII PN -CA1 PV synapses and may be a promising target for therapeutic interventions to treat disease progression.
The present study analyzed task performance in a hidden version of the Morris Water Maze test and revealed that AD mice exhibit spatial learning and memory defects beginning at 6 months of age Langston et al., 2010). Previous studies have reported that AD mice show deficits in task performance at 9 months of age  or after 14 months of age (Holcomb et al., 1999). However, several other studies have described the presence of behavioral deficits in AD mice as early as 3-6 months of age (King and Arendash, 2002;Lindner et al., 2006;Fritsch et al., 2010). The discrepancies among these studies may be due to differences in the experimental paradigms and in the genetic backgrounds of the mice used in the different laboratories. For example, some studies used a training schedule with 9 trials per day for 6 d, whereas others applied 10 trials per day for 4 d. The genetic background of the AD mice also affects the outcome of behavioral tests. In the present study, the AD mice and the nontransgenic controls were housed under the same conditions and were derived from the same litters; therefore, these parameters should not affect our conclusion that the activation of DAPK1 in ECII PN contributes to impairments in spatial learning and memory in AD mice.