Neuron-targeted caveolin-1 overexpression attenuates cognitive loss and pathological transcriptome changes in symptomatic Alzheimer’s disease models

SynCav1 gene delivery to early symptomatic PSAPP mice (6 m) preserved hippocampal-dependent learning and memory

To evaluate whether SynCav1 delivery at the symptomatic stage of AD attenuates cognitive deficits, the FC test was conducted at 12 m to assess learning and memory (Fig. 1a–c). Given our previous findings demonstrating that SynCav1 improves memory recall in normal WT mice,¹⁶ age and sex-matched WT-SynCav1 groups were omitted from the present study to reduce experimental redundancy. As shown in Fig. 1d, the WT-Sham group exhibited a rapid acquisition of freezing behavior over the first four tones on Day 1, ultimately reaching a plateau by tone 5. In comparison, both PSAPP-Sham and PSAPP-SynCav1 groups showed significantly fewer freezing behavior during the first four tones. Notably, at tone five, while PSAPP-SynCav1 group displayed a similar percent freezing time compared to the WT-Sham group (62.87% vs. 69.39%; p-value = 0.17), the PSAPP-Sham group still exhibited significantly fewer freezing behavior versus WT (54.47% vs. 69.39%; p-value < 0.01), indicating impaired acquisition ability in the 12 m PSAPP-Sham group. On Day 2, while PSAPP-Sham mice exhibited a significantly decreased percent freezing time versus WT-Sham mice (44.77% vs. 65.07%; p-value = 0.001), PSAPP-SynCav1 mice displayed a percent freezing similar to WT-Sham mice (58.60% vs. 65.07%; p-value = 0.28), indicating preserved hippocampal-dependent contextual memory recall (Fig. 1e). Assessment of cue-dependent memory recall on Day 3 revealed that both PSAPP-Sham and PSAPP-SynCav1 mice exhibited a decreasing trend in freezing behavior compared to WT-Sham mice. No significant difference was detected between PSAPP-SynCav1 and PSAPP-Sham groups (Fig. 1f).

OF test was also performed to assess locomotion and anxiety-like behavior (Fig. 1g–i). No significant difference in center dwelling time was detected among the three groups. Analysis of velocity and moving distance showed that both PSAPP-Sham and PSAPP-SynCav1 mice exhibited hyperlocomotion compared to WT-Sham mice, albeit no difference was observed between the two AD groups. These data indicate SynCav1 hippocampal delivery was not accompanied by adverse side effects related to anxiety or general locomotion.

SynCav1 preserved a WT-like transcriptome profile in PSAPP mice

Given the robust neuroprotective effect of SynCav1 on hippocampal-dependent functions observed in PSAPP mice, bulk RNA-seq was performed to further dissect the effect of SynCav1 on the transcriptomic profile of PSAPP mice. Principal component (PC) analysis determined that the AD (PSAPP)-SynRFP group distinctly separated from the WT control group, indicating a unique global gene expression profile that reflects the underlying AD pathology. In contrast, transcriptional profiles of AD-SynCav1 (i.e., PSAPP-SynCav1) and WT mice demonstrated a substantial overlap in the PC1/PC2 dimensions that accounted for 37.6% and 15.1% total gene expression variation, respectively (Fig. 2a). This overlay indicates a normalization of the gene expression patterns towards a non-diseased state. The phenotypic identity of mice in AD-SynRFP, AD-SynCav1, and WT groups was confirmed by mapping raw reads to the human CAV1 (GenBank ID 857), amyloid beta precursor protein (APP, GenBank ID 351) and presenilin 1 (PSEN1, GenBank ID 5663) mRNA sequences (Fig. 2b). Hierarchical clustering analysis of the top genes ranked by decreasing variance uncovered clusters of distinctly expressed and co-expressed genes across groups and biological replicates (Fig. 2c, d). Overexpression of Cav-1 was listed in the top genes as expected.

Differentially expressed genes (DEGs) were determined using thresholds of Padj <0.1. (supplementary Table S2). Accordingly, in the AD-SynCav1 vs. AD-SynRFP comparison group, 133 and 337 DEGs were estimated to have higher and lower expression, respectively (Fig. 2e). 61 and 4 DEGs exhibited higher and lower expression, respectively, in AD-SynCav1 compared to WT group (Fig. 2f). Comparison between AD-SynRFP and WT groups revealed 377 upregulated and 109 downregulated genes, respectively.

PSAPP mice exhibit altered neuronal/synaptic signatures that were restored with SynCav1

Protein-coding and non-coding RNA (ncRNAs) DEGs annotated in the GO database and fitting the standard statistical criteria (P < 0.05, |log2FC | > 0.58) in each WT vs. AD-SynRFP, AD-SynCav1 vs. AD-SynRFP and AD-SynCav1 vs. WT comparison groups were used for enrichment analyses in clusterProfiler.¹⁹ Signaling pathways from Kyoto Encyclopedia of Genes and Genomes (KEGG) and GO biological processes (BP) terms were ranked based on false discovery rate estimations (Q-values < 0.05), and selected relevant pathways are shown (Fig. 3). The list of predicted GO and KEGG pathways is included in Supplementary Tables S3 and S4. KEGG pathway analysis among 3 groups revealed that multiple dysregulated pathways in AD were reversed in the SynCav1-treated group (Fig. 3a). Interestingly, several genes implicated in neurodegenerative disease pathways, including AD (mmu05010), Huntington’s Disease (mmu05016), Parkinson’s Disease (mmu05012), ALS (mmu05014), pathways of neurodegeneration - multiple diseases (mmu05022) were significantly upregulated in AD compared to WT (pink arrows). Conversely, the AD-SynCav1 group suggested significant downregulation of these neurodegenerative disease pathways. A similar trend was observed in pathways related to synaptic function, as multiple dysregulated pathways involving glutamatergic synapses (mmu04724), GABAergic synapses (mmu04727), dopaminergic synapses (mmu04728), serotonin synapse6) (mmu04724, and cholinergic synapses (mmu04725) in AD were reversed in the AD-SynCav1 group (blue arrows).

As shown in Fig. 3b, GO biological processes analysis indicated that the AD-SynCav1 group exhibits higher levels of “learning or memory” (GO:0007611), “Cognition” (GO:0050890), “learning” (GO:0007612), and “memory” (GO:0007613) (red arrows). Furthermore, the AD-SynCav1 group also exhibits higher levels of synaptic-related pathways (blue arrows), including “synaptic structure and activity” (GO:0050803), “synaptic vesicle cycle” (GO:0099504), “neurotransmitter secretion” (GO:0007269), etc. Since synaptic loss is one of the characteristic anatomical hallmarks that correlate with cognitive deficits in AD patients, these results suggest that neuroprotective effects afforded by SynCav1 may occur by preserving or augmenting functional synaptic plasticity. As shown in Fig. 3c, tubulin binding, structural constituent of myelin sheath, microtubule binding, cell adhesion molecule binding, and actin binding were identified as the top 5 GO molecular functions. Synaptic membrane, sperm flagellum, motile cilium, 9 + 2 motile cilium, and collagen−containing extracellular matrix were identified as the top 5 GO cellular components (Fig. 3d).

SynCav1 gene delivery to mid-symptomatic APPKI mice preserved learning and memory

To further establish the applicability of SynCav1 therapy in the treatment of AD, we tested the effect of SynCav1 in a more clinically relevant APP knock-in mouse model that expresses physiological level of humanized mutant APP. Interestingly, a significant decrease in Cav-1 protein expression was detected in hippocampal tissue from APPKI mice at 8 m (Fig. 4a). We thus delivered SynCav1 to 8-month-old APPKI mice, which already exhibit a wide spectrum of AD-associated phenotypes, including memory deficit, amyloid-β deposits, and widespread neuroinflammation,²⁰ and conducted behavior tests at 12 months of age (Fig. 4b).

On Day 1 of FC, both APPKI-Sham mice and APPKI-SynCav1 mice showed significantly lower percent freezing time versus WT-Sham mice, suggesting a temporal limitation to the neuroprotective effects of SynCav1 on hippocampus-dependent associative learning in this AD model (Fig. 4c). On Day 2, a significant decrease in freezing behavior was detected in APPKI-Sham mice versus WT-Sham mice (41.62% vs. 59.29%; p-value < 0.001). In contrast, APPKI-SynCav1 exhibited significantly more frequent freezing behavior versus APPKI-Sham (50.03% vs. 41.62%; p-value = 0.01) (Fig. 4d), indicating that SynCav1 preserved hippocampal-dependent contextual memory recall in APPKI mice. On Day 3, both APPKI-Sham and APPKI-SynCav1 groups displayed significantly fewer freezing behavior versus the WT-Sham group, indicating deficits in non-hippocampal-dependent cue memory (Fig. 4e), an expected result since cued memory involves multiple brain regions, such as the amygdala and prefrontal cortex, that did not receive SynCav1.

Given that sex is a crucial variable in the AD-affected patient population, we also assessed the effect of SynCav1 therapy in female APPKI mice. Both APPKI-Sham and APPKI-SynCav1 groups exhibited decreased freezing on Day 1 (Fig. 4f), consistent with what was observed in male APPKI cohorts. On Day 2, female APPKI-SynCav1 mice displayed significantly greater contextual freezing behavior versus APPKI-Sham mice (51.16% vs. 37.06%; p = 0.03), demonstrating preserved hippocampal-dependent contextual memory recall (Fig. 4g). On Day 3, both APPKI groups displayed significantly fewer freezing behavior compared to the WT-Sham group (Fig. 4h).

OF test was performed to evaluate general locomotion and anxiety-like behavior. In the male cohorts, both APPKI-Sham and APPKI-SynCav1 mice exhibited significantly lower moving distance and velocity compared to the WT-Sham group; no significant difference was observed between APPKI-Sham and APPKI-SynCav1 (Fig. 4i). No significant difference was detected in time spent in arena center among all three groups. In the female cohorts, no difference in total distance traveled, velocity, and time spent in the arena center across all three groups (Fig. 4j). Overall, we have demonstrated that SynCav1 hippocampal delivery consistently preserved hippocampal-dependent contextual memory recall in two AD transgenic animal models without inducing side effects on general motor function and anxiety-related behaviors.

SynCav1 increased p-CaMKII and p-CREB expression in primary cortical neurons

In consideration of the robust cognitive functional improvement and upregulation in learning or memory-related pathways in the AD-SynCav1 group relative to AD group as identified by enrichment of GO for biological processes (Fig. 5a), we further assessed the effect of SynCav1 on synaptic plasticity signaling using primary cortical neurons. IB revealed that the expression of phosphorylated calcium/calmodulin kinase II (p-CaMKII) and phosphorylated cyclic-AMP response element-binding protein (p-CREB) (Fig. 5b, c), two molecular signatures of neuronal activity known to modulate synaptic plasticity and contribute to long-term memory formation, was significantly upregulated in SynCav1-transfected neurons compared to SynRFP-transfected control neurons.

SynCav1 preserved the expression of activity-dependent neuroprotective protein (ADNP)

Further analysis of enrichment of GO pathways involved in learning, memory, and cognition revealed ADNP as one of the genes with priority ranking in Hierarchical clustering analysis. Since ADNP has been shown to afford neuroprotective effects via stabilizing microtubules, protecting axonal transport, and attenuating oxidative stress,²¹ we thus validated the expression of ADNP in hippocampal tissue by IB. SynCav1 injection raised hippocampal Cav-1 expression by 8-fold (Fig. 6a, b). Compared to PSAPP-Sham mice, a significantly higher ADNP protein expression was detected in the hippocampus of PSAPP-SynCav1 mice, results consistent with findings from RNA sequence analysis. A downward trend (p = 0.12) in ADNP expression, albeit not significant, was observed in PSAPP-Sham mice compared to WT mice. No significant difference was found in full-length (FL) APP between PSAPP-Sham and PSAPP-SynCav1 mice.

In APPKI mice, SynCav1 injection resulted in an approximately 5-fold increase in hippocampal Cav-1 expression (Fig. 6c–e). Similar to what was observed in PSAPP mice, while a significant decrease in ADNP expression was detected in the hippocampus of both male and female APPKI-Sham mice compared to the WT-Sham group, APPKI-SynCav1 mice exhibited significantly higher ADNP protein expression compared to APPKI-Sham mice, confirming that SynCav1 preserved ADNP expression in an additional AD mouse model. No significant change in FL-APP expression was detected between APPKI-Sham and APPKI-SynCav1 groups.

SynCav1 preserved ADNP expression by maintaining MLR-localized PAC1R expression

Since ADNP expression could be upregulated by PAC1R activation by pituitary adenylate cyclase-activating peptide (PACAP), and the activation of this signaling pathway has been shown to exert neuroprotective effects in cerebral ischemia and hemorrhagic injury,^22–24 we examined PAC1R expression in hippocampal homogenates and MLRs-containing subcellular fractions. No significant difference in PAC1R expression was detected at the whole-cell level among all three groups (Fig. 7a, b). However, IB of the MLRs-containing fractions revealed a significantly greater expression of PAC1R in the APPKI-SynCav1 group compared to the APPKI-Sham group, which showed significantly decreased PAC1R expression in the MLRs versus the WT-Sham group (Fig. 7c–e). These findings imply that the WT-level ADNP expression in SynCav1-treated mice was likely due to the preservation of PACAP/PAC1R signaling.

Animals

All mice (PSAPP-transgenic (Tg) (APPSwePS1d9; The Jackson Laboratory, Bar Harbor, ME, USA, APPKI (App^{NL-G-F/NL-G-F}, Riken Institute, Japan)) were treated in compliance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health, Bethesda, MD, USA). Transgene (Tg) negative mice from the PSAPP strain were used as controls (WT), and C57/BL6J mice were used as control for APPKI groups. Mice were housed under normal conditions with a sufficient supply of water and food. PSAPP mice at 6 months of age (6 m) were selected for SynCav1 delivery, considering that impaired long-term memory was already detected by this age both in our hand (Supplementary Fig. 1a–c) and other groups.²⁵ Tg negative and PSAPP mice were allocated to 3 groups randomly: WT-Sham, PSAPP-Sham, and PSAPP-SynCav1 to receive either hippocampal stereotactic injections of AAV9-SynCav1 viral vector or sham surgery. Sham surgery involves creating needle paths using the same coordinates at 6 m. For transcriptome analyses, either SynRFP or SynCav1 was delivered to the hippocampus to correct for potential AAV9 effects on viral infection (e.g., neuroinflammatory response). APPKI mice (8 m) were divided to 2 groups randomly: APPKI-Sham and APPKI-SynCav1; WT control mice (8 m) received sham surgery. At 12 m, open field (OF) and fear conditioning (FC) neurobehavioral tests were performed to evaluate general behavior and cognitive function, respectively, for both PSAPP and APPKI mice. At the conclusion of all behavior testing, mice were humanely euthanized, and the hippocampus was collected for biochemical and histological analysis (Fig. 1a). All animal use protocols (#20-030) were approved by the Veterans Administration San Diego Healthcare System Institutional Animal Care and Use Committee (IACUC) before procedures were performed.

Stereotactic injection

Hippocampal delivery of AAV9-construct was performed on PSAPP (6 m) and APPKI mice (8 m). Briefly, mice were mounted onto a stereotaxic frame under anesthesia (2% isoflurane). Bilateral burr holes were made with a hand-held micro drill, and hippocampal injection was done using a 33-gauge, 10-μL gas tight syringe (Hamilton, Reno, Nevada) controlled by Injectomate (RWD, Shenzhen, China). A total of 3 μL of adeno-associated virus serotype 9 (AAV9) (viral titer: 2 × 10¹⁰ genome copies (g.c.)/μL) containing SynCav1 was injected bilaterally into hippocampal region over 180 seconds at three locations in each hemisphere (0.5 μL/site, 1st site: AP: 1.80 mm, Lat: 1.50 mm, DV: 1.7 mm; 2nd site: AP: 2.30 mm, Lat: 2.20 mm, DV: 1.75 mm; 3rd site: AP: 3.00 mm, Lat: 3.00 mm, DV: 3.00 mm) with 1 minute indwelling time (Fig. 1b). Sham surgery entails creating needle paths using the same coordinates.

Behavior tests

Statistical analysis

All data were analyzed using GraphPad Prism 8 (GraphPad Software, La Jolla, California, USA). For comparison between two groups, data was checked for normal distribution and analyzed by either Student’s t test. For comparison between three groups, data was analyzed using either One-way ANOVA or Two-way ANOVA followed by Bonferroni’s post hoc analysis or LSD multiple post hoc analysis as described. All data is presented as mean ± standard error of the mean (SEM), group size for each independent experiment was described in the figure legend and significance was assumed when p < 0.05. Experimental groups were blinded to the observer, and the code was broken for final analysis.