14-3-3θ Does Not Protect against Behavioral or Pathological Deficits in Alzheimer’s Disease Mouse Models

Abstract Alzheimer’s disease (AD) is characterized by progressive cognitive impairment associated with synaptic dysfunction and dendritic spine loss and the pathologic hallmarks of β-amyloid (Aβ) plaques and hyperphosphorylated tau tangles. 14-3-3 proteins are a highly conserved family of proteins whose functions include regulation of protein folding, neuronal architecture, and synaptic function. Additionally, 14-3-3s interact with both Aβ and tau, and reduced levels of 14-3-3s have been shown in the brains of AD patients and in AD mouse models. Here, we examine the neuroprotective potential of the 14-3-3θ isoform in AD models. We demonstrate that 14-3-3θ overexpression is protective and 14-3-3θ inhibition is detrimental against oligomeric Aβ-induced neuronal death in primary cortical cultures. Overexpression of 14-3-3θ using an adeno-associated viral (AAV) vector failed to improve performance on behavioral tests, improve Aβ pathology, or affect synaptic density in the J20 AD mouse model. Similarly, crossing a second AD mouse model, the AppNL-G-F knock-in (APP KI) mouse, with 14-3-3θ transgenic mice failed to rescue behavioral deficits, reduce Aβ pathology, or impact synaptic density in the APP KI mouse model. 14-3-3θ is likely partially insolubilized in the APP models, as demonstrated by proteinase K digestion. These findings do not support increasing 14-3-3θ expression as a therapeutic approach for AD.


Introduction
Alzheimer's disease (AD) is the most common neurodegenerative disease, with progressive neuronal loss leading to gradual cognitive decline and ultimately death. With the anticipated rise in AD cases in the next several decades, the expected societal burden and financial costs of AD are skyrocketing, with predicted health care costs for AD reaching $1.2 trillion in 2050 (Alzheimer's Association, 2014). In addition to neuronal loss in the hippocampus and frontotemporal cortex, AD is characterized pathologically by two major hallmarks, b -amyloid (Ab ) plaques and tau tangles. However, cognitive impairment correlates more strongly with synaptic dysfunction and spine loss than with plaque or tangle burden (DeKosky and Scheff, 1990;Terry et al., 1991;Hardy and Selkoe, 2002;Selkoe, 2011;Pozueta et al., 2013;Tu et al., 2014;Boros et al., 2017).
Taken together, this evidence suggests that 14-3-3u dysfunction contributes to cognitive deficits in AD. In this study, we used primary neuronal cultures and two AD mouse models, the J20 APP transgenic model and the App NL-G-F knock-in (APP KI) mouse, to test the impact of 14-3-3u in AD. While we found that 14-3-3u reduced Abinduced neuron loss in culture, 14-3-3u failed to reduce Ab plaque load, alter synaptic density, or improve cognitive deficits in the AD mouse models.

Mice
All mice used in these studies followed the guidelines of the National Institutes of Health (NIH) and University of Alabama at Birmingham Institutional Animal Care and Use Committee (IACUC), and all work performed was approved by University of Alabama at Birmingham IACUC. Equal numbers of male and female mice were used for all studies. Mice were group housed under 12/ 12 h light/dark cycle. The 14-3-3u line used expressed human 14-3-3u under the neuronal promoter Thy1.2 (Lavalley et al., 2016;Wang et al., 2018;Underwood et al., 2021). The difopein line used expressed difopein-enhanced yellow fluorescent protein (eYFP) under the promoter Thy1.2 and was initially obtained from Yi Zhou at Florida State University (Qiao et al., 2014). To obtain both transgenic and littermate controls for the primary culture experiments, both hemizygous 14-3-3u and difopein mice were crossed separately with C57BL/6J mice from The Jackson Laboratory (catalog #000664; RRID: IMSR_ JAX:000664). J20 transgenic mice carry both the APP Swedish (KM670/671NL) and APP Indiana (V717F) mutations and were on a C57BL6/J background (Mucke et al., 2000). Homozygous App NL-G-F/NL-G-F (APP KI) mice (Saito et al., 2014) were provided by the Riken BRC through the National Bio-Resource Project of the MEXT, Japan. Hemizygous 14-3-3u transgenic mice (14-3-3u tg) were crossed with homozygous APP KI mice (App NL-G-F/NL-G-F ) to create App WT/NL-G-F /14-3-3u tg and App WT/NL-G-F /WT. These mice were then crossed with each other to yield, among other genotypes, the desired App NL-G-F/NL-G-F /14-3-3u tg and App NL-G-F/NL-G-F /WT mice. Crossing these two mice with each other produced the App NL-G-F/NL-G-F /14-3-3u tg (APP KI/14-3-3u ) and App NL-G-F/NL-G-F /WT (APP KI) mice used in experiments, without the additional possible genotypes. Control mice (nTg) without either the mutant APP or the overexpressed 14-3-3u were created by crossing a heterozygous APP KI mouse (App WT/NL-G-F ) with a nTg C57BL/6 mouse.

Ab oligomer (Abo) preparation
Ab (rPeptide, A-1163-2) was reconstituted in HFIP (Sigma-Aldrich, 105228) and aliquoted. After full evaporation of HFIP in fume hood, aliquots were stored at À80°C until ready for use. One day before experiments, an Ab aliquot was resuspended to 100 mM. The Ab was first resuspended in a small volume in sterile DMSO (Sigma D2650) by vortexing and brought to the final volume using HBSS (Invitrogen14,175-095). After pipette trituration, the Ab was sonicated on high for 15 cycles of 45 s on, 15 s off. The Ab was then left overnight at 4°C to form oligomers.

Ab toxicity assay
Hippocampal or cortical regions were dissected from male and female postnatal (P)0 mice and incubated at 37°C for 25 min in papain. Cells were then washed with neurobasal-A media containing B-27 supplement, Glutamax, and 5% FBS and then triturated using fire polished glass pipettes. Cells were pelleted by centrifugation at 700 rpm for 5 min and were resuspended in neurobasal-A media containing B-27 supplement, Glutamax, and 5% FBS and plated on 18 mm poly-D-lysine-coated coverslips. After 12-16 h, media were replaced by neurobasal-A media containing B-27 supplement, Glutamax, and Arabinose C at 6 mm. Fifty percent media changes were made every 7 d. After 14 d, 1 or 5 mM of Ab oligomers (Ab o) or vehicle control were added. After 24 h of Ab o treatment, cells were incubated with ethidium homodimers to stain dead cells and Hoechst dye to stain all cells, and then mounted onto slides using VectaShield (Vector Laboratories). Total cells stained with both ethidium homodimers and Hoechst were then counted and a percentage of dead cells calculated.

Behavior
All behavior tests were performed with approximately equal numbers of male and female mice.

Open field
Mice were placed in a 48 Â 48-inch open arena with clear plexi-glass walls and allowed to freely move for 4 min while being recorded using EthoVision software.
Video was then analyzed for overall velocity and distance moved across all planes (vertical and horizontal).

Elevated plus maze (EPM)
Mice were placed in the center of plus shaped maze elevated above the floor. Two oppositely placed arms of the maze are open and the remaining two arms are closed. Mice were allowed to freely explore the maze for 4 min while being recorded using EthoVision software. Video was then analyzed to calculate the percentage of time that mice spent in the open arm.

Morris water maze (MWM)
Mice were placed in a pool of water with a hidden platform submerged under the surface of the water. During the acquisition phase of the test, the time taken for the mice to find the platform (latency to platform) was recorded for four trials daily for 5 d. Immediately following the final trial on day 5 of the acquisition phase, a probe trial was conducted where the platform was removed. The percentage of time spent in the quadrant where the platform had been located (target quadrant) was then calculated. All trials were recorded and analyzed using EthoVision software.

Passive avoidance
The experiment was performed using a two-compartment box in which one chamber was darkened, and a grid floor of stainless-steel bars was connected to an internal shock source (Gemini II Avoidance System). On the training day, each mouse was placed in the lighted chamber with the interchamber door closed. After 60 s the door was opened and when the mouse crossed to the darkened side the door was closed and a 0.5 mA Â 2 s foot shock was administered. Mice were left in the dark chamber for 30 s before being returned to their home cage. Mice were tested 24 h later and the protocol was repeated but without application of a foot-shock, with a cutoff time of 5 min. For both training and testing, the latency was measured as the time when all four paws entered into the darkened chamber.

Ab plaque load quantification
Every fourth section through the hippocampus was stained for Ab , using the TSA Plus amplification kit (Akoya Biosciences, NEL763001KT). Sections were de-waxed and rehydrated and antigen retrieval was done by autoclaving in citrate buffer. Sections were quenched with 0.3% hydrogen peroxide in methanol before blocking and overnight incubation with primary antibody (mouse anti-Ab 82E1, IBL 10 326). After washing, sections were incubated with goat anti-mouse IgG biotinylated secondary antibody (Vector Laboratories, BA-9200). Sections were then incubated in HRP-conjugated streptavidin (Thermo Scientific) and finally with the FITC-conjugated Tyramide. Slides were coverslipped using ProLong Diamond Antifade mounting solution (Thermofisher Scientific) and imaged using an Olympus BX51 epifluorescence microscope. Images were then quantified using the 'analyze particles' function in ImageJ to calculate percent plaque load in both hippocampus and cortex.

Synapsin quantification
To estimate synaptic density, paraffin embedded sections were stained for the presynaptic marker synapsin. Paraffin embedded sections were stained as described for the Ab immunohistochemistry methods using a rabbit synapsin primary antibody (Invitrogen 51-5200), followed by the TSA Plus amplification kit as described above. 60Â images of each hippocampal section were acquired using a Nikon Eclipse Ti2 confocal microscope. The area positive for synapsin staining was quantified using ImageJ software. Images were converted to 8 bit in ImageJ, and the threshold was set with a lower limit of 15 and upper limit of 255. The area fraction and total area of the image was obtained through the "measure" feature of ImageJ. Any blank spaces in the image because of tears in the tissue were then circled as ROIs and the total area of the empty space was obtained using the "measure" feature again. The empty space area was subtracted from the initial total area of the image, and the percent area positive for synapsin staining was calculated. For each mouse, two images from each of three to four hippocampal sections were analyzed and averaged.

Proteinase K digestion
Sections were dewaxed and rehydrated and then treated for 10 min at 37°C with 20 mg/ml of proteinase K diluted in TBS or TBS only. Antigen retrieval was done by autoclaving in citrate buffer and sections were quenched with 0.3% hydrogen peroxide in methanol. Sections were then blocked in 5% NGS with 0.1% Triton X-100 and incubated overnight with primary antibody (mouse 14-3-3u ; Santa Cruz SC-69720; RRID: AB_2218224). After washing with TBS, sections were incubated with Cy3-conjugated goat anti-mouse secondary antibody. Slides were coverslipped using ProLong Diamond Antifade mounting solution (Thermofisher Scientific) and imaged using a Nikon Eclipse Ti2 confocal microscope.

Statistical analysis
All statistical analyses and graphing were performed in Prism 9.2 (GraphPad Software, Inc., La Jolla, CA). Details of statistical tests are outlined in Table 1.
14-3-3h does not reduce cognitive decline in AD mouse models We next examined the effects of 14-3-3u overexpression in two different AD mouse models: (1) the J20 APP mouse (Mucke et al., 2000), and (2)   1. 14-3-3u reduces oligomeric Ab toxicity in primary cortical cultures. A, Immunocytochemistry for exogenous HA-tagged 14-3-3u in nontransgenic (nTg) and 14-3-3u transgenic mouse neuronal cultures. Scale bar: 100 mm. B, Representative images of primary cortical neurons treated with oligomeric Ab from nontransgenic or 14-3-3u mice. Ethidium D (EthD) labels the nuclei of dying cells, while Hoechst 33342 stains the nuclei of all cells. Scale bar: 100 mm. C, Immunocytochemistry for eYFP-difopein in nontransgenic and difopein transgenic mouse neuronal cultures. Scale bar: 100 mm. D, Representative images of primary cortical neurons treated with oligomeric Ab from nontransgenic or difopein mice. Ethidium D labels the nuclei of dying cells, while Hoechst 33342 stains the nuclei of all cells. Scale bar: 100 mm. E, Western blotting for 14-3-3u levels in Triton X-100 soluble fractions from hippocampal (H) and cortical (C) cultures from nontransgenic and 14-3-3u mice. F, Quantification of cell death in primary cortical neurons from nTg or 14-3-3u littermate mice treated with Ab oligomers for 24 h. n = 6 per condition; **p 0.01, ****p 0.0001 (Tukey's multiple comparison test). Error bars represent standard error of the mean (SEM). G, Quantification of cell death in primary hippocampal neurons from nTg or 14-3-3u littermate mice treated with Ab oligomers for 24 h. n = 7 per condition. Error bars represent SEM. H, Quantification of cell death in primary cortical neurons from nTg or difopein littermate mice treated with Ab oligomers for 24 h. n = 4 per condition, **p 0.01, ****p 0.0001 (Tukey's multiple comparison test). Error bars represent SEM. I, Quantification of cell death in primary hippocampal neurons from nTg or difopein littermate mice treated with Ab oligomers for 24 h. n = 4 per condition. Error bars represent SEM. 2005). The newer APP KI model expressing human APP carrying the Swedish, Beyreuther/Iberian (I716F), and Arctic (E22G) mutations also demonstrates memory deficits and Ab plaque deposition (Saito et al., 2014;Masuda et al., 2016).
We used both viral-mediated overexpression and transgenic means to test the impact of 14-3-3u overexpression in vivo. We first used stereotactic injections of an AAV vector to overexpress 14-3-3u in the hippocampi of nTg and J20 mice. Male and female 8-to 11-week-old mice were injected with either AAV-GFP or AAV-14-3-3u /GFP into the hippocampi bilaterally. We confirmed expression of GFP or V5-tagged 14-3-3u in the hippocampi of AAVinjected mice at the immediate conclusion of the behavioral experiments, approximately six months after viral injection ( Fig. 2A-D). At two and six months after viral injection, we performed behavioral tests, including the open field, EPM, and MWM tests. The J20 mice have been previously shown to demonstrate behavioral alterations in these tests with increased exploratory movement on the open field test, altered anxiety on the EPM, and impaired memory on the MWM (Chin et al., 2005;Kobayashi and Chen, 2005;Ognibene et al., 2005;Cheng et al., 2007;Roberson et al., 2007;Meilandt et al., 2009;Harris et al., 2010).
Behavioral tests were repeated on the same cohort of mice at six months after AAV injection, when the mice were eight to nine months old, though the sample sizes for these later behavioral tests were limited because of J20 mouse mortality. While not statistically significant, the 14-3-3u -injected J20 mice showed a higher mortality rate than the GFP-injected J20 mice (p = 0.201; log-rank/Mantel-Cox test; Fig. 2I). Premature mortality rates around 15% have been described in the J20 mouse line . We observed that 2 out of 10 (20%) GFP-injected J20 mice died compared with 5 out of 11 (45%) 14-3-3u -injected J20 mice by the conclusion of the experiment (Fig. 2I).
We also examined the effect of 14-3-3u in the APP KI model by crossing 14-3-3u transgenic mice with APP KI mice. Unlike the J20 mice, we did not observe any mortality of these mice by the end of the experiment (Fig. 3B). As APP KI mice demonstrate much more subtle cognitive changes at later time points, we tested these mice at eight to nine months by open field, elevated zero maze, and the passive avoidance test (Masuda et al., 2016;Saito et al., 2014). The APP KI mice do not demonstrate deficits on the MWM, and thus we did not perform this test (Saito et al., 2014). We examined nTg (App WT/WT / WT), APP KI (App NL-G-F/NL-G-F /WT), and APP KI/14-3-3u (App NL-G-F/NL-G-F /14-3-3u tg) mice on these behavioral tests. After the behavior was concluded and the mice were killed, we confirmed exogenous HA-tagged 14-3-3u expression in the cortex and hippocampus of APP KI/14-3-3u mice, which was not observed in APP KI mice (Fig. 3A). NTg, APP KI, and APP KI/14-3-3u mice showed no difference in distance traveled in the open field (one-way ANOVA: F (2,34) = 0.9122, p = 0.4112; Fig. 3C). APP KI and APP KI/14-3-3u mice spent more time in the open arm in the elevated zero maze compared with the nTg mice, indicating decreased anxiety, but there was no difference between APP KI and APP KI/14-3-3u mice (one-way ANOVA: F (2,31) = 11.17, p = 0.0002; Fig. 3D). On the passive avoidance test, none of the three groups showed any delay in crossing to the dark side of the shuttle box on day 2, and there was no difference between the groups (one-way ANOVA: F (2,29) = 0.1705, p = 0.8441; Fig. 3E). Of note, the passive avoidance test may have been impacted by an unexpected and uncontrollable alarm that randomly went off in the animal facility during housing of the APP KI cohort. As with the J20 mice, 14-3-3u overexpression did not rescue behavioral changes in the APP KI mice.

14-3-3h does not affect Ab plaque deposition in vivo
Immediately on the conclusion of the last behavioral time point, both cohorts of mice were killed, and their brains were examined for Ab plaque load using immunofluorescent staining. As expected, the J20 mice had Ab plaques in the hippocampus, whereas Ab plaques were not observed in the nTg controls (Fig. 4A). There was no difference in hippocampal Ab plaque load between the GFP-injected and 14-3-3u -injected J20 mice (unpaired, two-tailed t test: t (9) = 0.09254, p = 0.9238; Fig. 4A). We did not analyze plaque load in the cortex of J20 mice as the AAV injection was directed to the hippocampi. Similarly, the APP KI mice had a large number of plaques in both the hippocampus and cortex, but there was no difference between the APP KI and APP KI/14-3-3u in either the hippocampus (unpaired, two-tailed t test: t (24) = 1.038, p = 0.3084; Fig. 4B) or cortex (unpaired, two-tailed t test: t (23) = 1.475, p = 0.1538; Fig. 4C). Based on these findings, 14-3-3u overexpression did not reduce Ab plaque burden in either the J20 or APP KI mouse models.

14-3-3h does not affect synaptic density in APP models in vivo
We also examined mouse brains for potential loss of synaptic density by staining for the presynaptic marker synapsin in both cohorts of mice. In the J20 mice, there was no difference in the area that stained positive for synapsin, regardless of genotype or injection type (two-way ANOVA: genotype F (1,23) = 0.1737, p = 0.6807; virus F (1,23) = 0.7492, p = 0.3957; interaction F (1,23) = 0.3085, p = 0.5839; Fig. 5A). Similarly, the APP KI mice showed no changes in synapsin, regardless of genotype (one-way ANOVA: F (2,31) = 0.4169, p = 0.6627; Fig. 5B). Immunostaining with the presynaptic marker vesicular glutamate transporter vGLUT1 and postsynaptic marker NR2A confirmed synaptic integrity in the brains of both mouse cohorts (Fig. 5). Based on these results, we concluded that 14-3-3u overexpression did not alter synaptic density in either cohort of mice.

APP KI mice show altered subcellular 14-3-3h distribution and solubility
Given the lack of protective effect for 14-3-3u overexpression in the APP models, we examined the subcellular 14-3-3u distribution in J20 and APP KI mice. In GFP-injected J20 mice and in APP KI mice, 14-3-3u immunoreactivity was distributed in a very intense rim within the cytoplasm of neurons in deep cortical neurons and hippocampal neurons compared with that seen in nTg mice (Figs. 2C,3A,6). A similar pattern was observed in the presence of 14-3-3u overexpression in 14-3-3u -injected J20 mice and in the APP KI/14-3-3u mice (Figs. 2D,3A,6). This pattern of staining in some neurons suggested possible insolubilization of 14-3-3u in J20 and APP KI mice, as has been observed in human AD brains and in AD mouse models (G. Xu et al., 2013;McFerrin et al., 2017). To test for insoluble 14-3-3u , brain sections were treated with proteinase K. While proteinase K digestion eliminated 14-3-3u immunoreactivity in nTg mice, 14-3-3u immunoreactivity, though reduced, was still observed in APP KI and APP KI/14-3-3u mice with proteinase K digestion (Fig. 6).

Discussion
In this study, we sought to study the effect of 14-3-3u in AD, using both primary cultured neurons and in vivo AD mouse models. In primary culture, we found that 14-3-3u overexpression ameliorated, while pan 14-3-3 inhibition exacerbated, Ab toxicity in cortical cultures. No clear effect was observed in hippocampal cultures, pointing to a potential regional effect of 14-3-3 modulation. Based on Figure 3. 14-3-3u overexpression does not modify behavior in APP KI mice. A, Representative images of HA immunostaining in the cortex and hippocampus of APP KI mice demonstrates expression of HA-tagged 14-3-3u in 14-3-3u -overexpressing mice (APP KI/ 14-3-3u ) but not mice without overexpression (APP KI). 14-3-3u immunostaining detects both endogenous and exogenous 14-3-3u expressed in APP KI and APP KI/14-3-3u mice. Scale bar: 100 mm for HA images and 50 mm for 14-3-3u images. B, Quantification of mortality for nontransgenic, APP KI, and APP KI/14-3-3u mice. C, Quantification of distance traveled in the open field task at eight to nine months of age. n = 10-16 mice per group. Error bars represent SEM. D, Quantification of time spent in the open arm on the EPM at eight to nine months of age. n = 10-15 mice per group; *p 0.05, ***p 0.001 (Tukey's multiple comparison test). Error bars represent SEM. E, Quantification of latency in crossing to the dark side on day 2 of the passive avoidance task. n = 8-14 mice per group. Error bars represent SEM. these data, we hypothesized that overexpressing 14-3-3u in AD mice would lead to an improvement in both the cognitive deficits and pathology. However, 14-3-3u overexpression failed to reduce cognitive decline or Ab pathology or affect synaptic density in either the J20 or APP KI models. In the case of the J20 mice, we injected AAV-14-3-3u into the hippocampus because of the difficulty of obtaining widespread cortical AAV Figure 5. 14-3-3u overexpression does not impact synaptic density in AD mice. A, Representative hippocampal images of synapsin, vGLUT1, and NR2A immunoreactivity and quantification of synapsin immunoreactivity in nontransgenic and J20 mice injected with either AAV-GFP or AAV-14-3-3u at 6 MPI. n = 5-9 per group. Error bars represent SEM. Scale bars: 50 mm for synapsin images and 10 mm for vGLUT1/NR2A images. B, Representative hippocampal images of synapsin, vGLUT1, and NR2A immunoreactivity and quantification of synapsin immunoreactivity in nontransgenic, APP KI, and APP KI/14-3-3u mice at eight to nine months of age. n = 9-16 per group. Error bars represent SEM. Scale bars: 50 mm for synapsin images and 10 mm for vGLUT1/NR2A images. 4. 14-3-3u overexpression does not reduce Ab plaque burden in AD mice. A, Representative hippocampal images and quantification of total plaque area of J20 mice injected with either AAV-GFP or AAV-14-3-3u at 6 MPI. n = 5-6 per group. Error bars represent SEM. Scale bar: 500 mm. B, Representative hippocampal images and quantification of total plaque area of APP KI and APP KI/14-3-3u mice at eight to nine months of age. n = 10-16 per group. Error bars represent SEM. Scale bar: 500 mm. C, Representative cortical images and quantification of total plaque area of APP KI and APP KI/14-3-3u mice at eight to nine months of age. n = 10-15 per group. Error bars represent SEM. Scale bar: 500 mm.
expression. Based on our in vitro data, targeting the hippocampus would not necessarily be sufficient to rescue behavior or pathology. However, boosting 14-3-3u expression in both the cortex and the hippocampus, as was done in our APP KI Â 14-3-3u transgenic cross, was also not sufficient to rescue behavioral or pathologic changes in the cortex or hippocampus.
While it was unexpected to see 14-3-3u protection in culture but not in vivo, this discrepancy likely reflects the vast differences in the environments of in vitro and in vivo experiments. In the in vivo system, other brain cell types, such as glia and astrocytes, interact with the neuronal population and could affect the findings. Alternatively, 14-3-3u may be protective only in the short term, such that the protective effects were observed only in culture but not in vivo at the time points examined. Additionally, our in vitro studies showed that the protective effect was limited to cortical cultures and not to hippocampal cultures. This finding suggests that any protective effect of 14-3-3u could be restricted to certain neuronal populations such that any global protection may not be seen in vivo. Further, we propose that 14-3-3u overexpression failed to protect against Ab toxicity in vivo because of potential insolubilization of 14-3-3u . We found altered patterns of 14-3-3u immunoreactivity in both the APP KI mice and the APP KI/14-3-3u mice, in which neurons showed intense, "rim-like" cytoplasmic staining in neurons in the cortex and hippocampus, compared with less intense, more diffuse immunoreactivity in nTg control mice. This immunoreactivity in both APP KI mice and the APP KI/14-3-3u mice was partially resistant to proteinase K digestion, suggesting a reduction in the solubility of 14-3-3u , even in the presence of 14-3-3u overexpression. We hypothesize that the presence of the APP mutations since conception in vivo may drive the insolubilization and consequent dysfunction of 14-3-3u and thus block its protective effect, even in the presence of overexpression. In contrast, in vitro Ab oligomers are added to neurons cultured from 14-3-3u mice that do not contain the APP mutations so that 14-3-3u is soluble at the time of Ab o treatment, as demonstrated by Western blotting for 14-3-3u using Triton X-100 soluble fractions from primary neurons (Fig. 1C). Because the neurons are only exposed to the Ab oligomers for a relatively short time, it is likely not long enough to lead to the insolubility of 14-3-3u ; and therefore, 14-3-3u is able to maintain its protective functions.
In the J20 mice, we examined cognitive function using the MWM. At both 2 and 6 MPI, the J20 mice showed a reduced ability to locate the platform, but there was no significance difference between J20 mice who received the AAV-GFP injection compared with the AAV-14-3-3u /GFP. At the earlier time point of 2 MPI, the 14-3-3u -injected J20 mice actually showed a trend of doing worse than the GFP-injected J20 mice. While no difference in water maze performance was noted at the 6 MPI time point between GFP-injected and 14-3-3u -injected J20 mice, increased mortality of the 14-3-3u -injected J20 mice by 6 MPI could have limited our ability to detect a small difference. However, we did not see an increase in mortality in the APP KI/14-3-3u , which were similarly impaired compared with APP KI mice, so it is unlikely that our lack of improvement was due solely to the poor health or increased mortality of the 14-3-3u -injected J20 mice. 14-3-3u showed no protection against cognitive decline and potentially worsened cognitive function and mortality in the J20 model.
One surprising result of our study was a possible trend toward higher mortality of the 14-3-3u -injected J20 mice compared with the GFP-injected mice. A total of 45% (5/ 11) of the 14-3-3u -injected mice died before our planned kill date, whereas only 20% (2/10) of the GFP-injected mice failed to survive. Of note, we had previously noted a high mortality rate of 14-3-3u /J20 double transgenic mice when we tried to cross the 14-3-3u transgenic mice with the J20 mice as an initial experimental approach. Because of the early mortality of double transgenic mice from this cross, we decided to switch to the AAV approach. One potential cause for this increased death within the 14-3-3u -overexpressing J20 mice could be seizures. The J20 mouse line, like most lines expressing high levels of Ab , has hippocampal hyperexcitability that leads to nonconvulsive seizures and higher rates of premature mortality . It is possible that the overexpression of 14-3-3u increased seizure activity. Several studies have implicated 14-3-3s in seizure activity, although most of the data has pointed to a protective effect of other 14-3-3 isoforms in seizures (Schindler et al., 2006;Murphy et al., 2008;Brennan et al., 2013). Reducing tau in J20 mice reduces the severity of seizure activity in these mice , and 14-3-3z has also been shown to interact with tau and regulate tau aggregation in vivo (Sadik et al., 2009a,b). Therefore, it is possible that 14-3-3u may interact with tau to affect seizure activity in the J20 mice. Finally, 14-3-3 proteins affect both neuronal morphology and glutamatergic receptor trafficking (Gohla and Bokoch, 2002;Gehler et al., 2004;Simsek-Duran et al., 2004;Chen and Roche, 2009;Ramser et al., 2010;Yoon et al., 2012;Qiao et al., 2014;Chung et al., 2015;Foote et al., 2015;Lavalley et al., 2016;Kaplan et al., 2017), such that any 14-3-3u -induced alterations in receptor trafficking or neuronal morphology could impact seizure activity and thus mortality.
In conclusion, we examined the effect of 14-3-3u on Ab -induced toxicity in several AD models. While protective against Ab o-induced toxicity in cortical cultures, 14-3-3u overexpression failed to rescue cognitive decline and pathologic changes in two APP mouse models. These findings do not support the idea of manipulating 14-3-3u as a therapeutic approach for AD.