Alzheimer’s disease (AD) is the leading cause of dementia (Alzheimer’s Association, 2021), affecting around 35 million individuals worldwide hence representing an important strain on health resources (Cummings et al., 2021). However, effective disease-modifying pharmacologic therapies for AD are not currently available (Tatulian 2022). AD is causally related to the accumulation of abnormally folded amyloid-β (Aβ) peptide and hyperphosphorylated tau (pTau) protein aggregates, leading to synapse loss, neurodegeneration, and progressive cognitive decline. The widely accepted amyloid cascade hypothesis posits that the accumulation of Aβ peptides in the brain parenchyma is a key event of AD pathogenesis. Furthermore, a large body of evidence now indicates that soluble Aβ oligomers, rather than their insoluble fibrillar aggregates, are responsible for the synapto- and neurotoxicity of the Aβ peptide (Walsh et al., 2002; Karran et al., 2011). The soluble Aβ oligomers initiate a cascade of pathological pathways that trigger dendritic and synaptic alterations, which ultimately lead to synaptic plasticity impairment, neurodegeneration, and AD dementia (Selkoe and Hardy 2016; Ricciarelli and Fedele 2017). However, the underlying molecular mechanisms are not fully understood hence preventing the development of effective treatments.

The Rho protein kinase (Rock) is a serine/threonine protein kinase whose activity is regulated by the small GTPase RhoA (Strassheim et al., 2019). Once active, Rock phosphorylates various cellular substrates involved in actin cytoskeleton dynamics, hence regulating cell adhesion, contraction, migration, and division, as well as survival (Strassheim et al., 2019). The predominant brain Rock isoform is Rock2, which phosphorylates cofilin to regulate actin cytoskeleton (Zhang et al., 2006; Newell-Litwa et al., 2015), dendritic spine morphology and synaptic plasticity (Hensel et al., 2015). We recently described that Rock2 accumulation and activation in neurons triggers dendrite disruption and synapse loss, leading to memory impairment (Bobo-Jimenez et al., 2017). Moreover, Rock2 expression resulted in lower dendritic spine density (Henderson et al., 2019), whereas Rock inhibition increases the number of thin spines and filopodia, which would enhance synapse formation and neuronal plasticity (Swanger et al., 2015; Cai et al., 2021). Interestingly, dendritic dystrophy (Cochran et al., 2014; Shi et al., 2020) and synapse loss (Terry et al., 1991; Scheff et al., 2007) have both been detected in early-stage AD brains and are correlated with cognitive decline. Moreover, Rock2 protein accumulates in the neurons of early-stage human AD brain and remain elevated throughout the disease progression (Herskowitz et al., 2013). In experimental models, enhanced Rock2 activity is associated with typical AD hallmarks, such as Aβ aggregation, tau hyperphosphorylation, neuroinflammation, synaptic damage and neuronal death (Gao et al., 2019; Cai et al., 2021). However, the molecular mechanism that leads to Rock accumulation and activation in AD is unknown.

Previous results from our group have shown that the activator of the E3 ubiquitin ligase anaphase-promoting complex/cyclosome (APC/C), Cdh1, is essential for dendritic network integrity, synaptic plasticity, and neuronal survival (Almeida et al., 2005; Maestre et al., 2008; Almeida 2012; Bobo-Jimenez et al., 2017). Recently, we found that APC/C-Cdh1 targets Rock2 for proteasomal degradation in neurons, and that the loss of Cdh1 causes Rock2 stabilization and activation, triggering dendrite disruption, and dendritic spine and synapse loss in the adult brain following neurodegeneration and cognitive impairment (Bobo-Jimenez et al., 2017). Here, we hypothesized whether the regulation of Rock2 by APC/C-Cdh1 is an important mechanistic event in AD. In summary, using in vitro and in vivo experimental mouse models, we found that Aβ25-35 oligomers induce cyclin dependent kinase-5 (Cdk5)-mediated Cdh1 phosphorylation leading to APC/C inactivation and neuronal apoptosis. Interestingly, through this mechanism, Aβ25-35 triggers Rock2 accumulation and activation in neurons, inducing memory loss that could be rescued by selectively inhibiting Rock2 activity. Thus, here we identify a novel Cdh1-Rock2 pathway that is involved in Aβ neurotoxicity, which may open a new avenue for the development of therapeutic strategies to combat cognitive impairment in AD.

Aβ-induced Cdh1 phosphorylation causes APC/C inactivation and neuronal apoptosis. To understand the potential role of the Cdh1-Rock2 pathway in Aβ neurotoxicity, we first analyzed the effect of Aβ on APC/C-Cdh1 activity. Cdh1 is the main activator of APC/C in neurons, where it is regulated by autoubiquitination, Cdk5-mediated phosphorylation and subcellular localization (Maestre et al., 2008; Almeida 2012). Treatment of neurons with oligomers of the amyloidogenic fragment Aβ25-35 time-dependently enhanced Cdh1 protein levels (Figure 1A) until 8 h of incubation, especially in the cytosol (Figure 1B); this effect was prevented by roscovitine, a cyclin dependent kinase-5 (Cdk5) inhibitor (Figure 1B), suggesting that Aβ25-35 might cause Cdh1 phosphorylation (Maestre et al., 2008; Veas-Perez de Tudela et al., 2015b; Fuchsberger et al., 2016). In accordance with our previous results (Lapresa et al., 2019), Aβ25-35 rapidly (2 h) activated Cdk5 (Figure 1C), resulting in a time-dependent phosphorylation of Cdh1 (Figure 1D and Supplementary Figure S1D). Since Cdh1 phosphorylation at Cdk5-cognate residues disrupts Cdh1 assembly from its APC/C complex core protein APC3, leading to APC/C-Cdh1 inactivation in neurons (Veas-Perez de Tudela et al., 2015b), we next assessed whether these effects could be mimicked by Aβ25-35 treatment. As shown in Figure 1E, APC3-Cdh1 interaction was disrupted by Aβ25-35, an effect that was rescued with roscovitine. Moreover, Aβ25-35 treatment caused APC/C inactivation, as revealed by the decreased ubiquitination of its well-known substrate, cyclin B1 (Zachariae et al., 1998), in a manner that could be prevented by roscovitine and Cdk5 knock down (siCdk5) (Figure 1F). Thus, Aβ25-35-induced Cdk5 activation promotes the accumulation of phosphorylated (inactive) Cdh1 and its disassembly from APC3, leading to APC/C-Cdh1 inactivation.

Given that the inactivation of APC/C-Cdh1 caused by Cdh1 phosphorylation promotes neuronal apoptosis (Maestre et al., 2008; Veas-Perez de Tudela et al., 2015b; Veas-Perez de Tudela et al., 2015a), we next evaluated the impact of Cdh1 in Aβ25-35 toxicity. As shown in Figure 2A, Aβ25-35-induced caspase 3 activation that was partially prevented by roscovitine. Then we investigated whether Cdh1 phosphorylation by Cdk5 impacted in Aβ25-35 neurotoxicity. To this end, a phosphomimetic (Cdh1-D) or a phosphodefective (Cdh1-A) form of Cdh1 were expressed in neurons followed by Aβ25-35 treatment. As previously described, Cdh1-D resulted in neuronal apoptosis (Maestre et al., 2008; Veas-Perez de Tudela et al., 2015b). In contrast, Cdh1-A prevented Aβ25-35-induced neuronal apoptosis (Figure 2B and Supplementary Figure S2A). Moreover, inhibition of APC/C using ProTAME and Cdh1 knock down (siCdh1) enhanced Aβ25-35-induced neurotoxicity, as revealed by the increased neuronal apoptosis and caspase 3 activation (Figures 2C,D and Supplementary Figure S2A–C). Thus, Cdk5-mediated Cdh1 phosphorylation triggers APC/C inhibition, which is important in Aβ25-35 neurodegeneration.

Aβ-induced APC/C- Cdh1 inactivation triggers Rock2 stabilization and activation leading to neuronal apoptosis. Next, we sought to identify the APC/C-Cdh1 target that is involved in Aβ25-35 neurotoxicity. We specifically focused on Rock2, given that it is a well-known APC/CCdh1 substrate (Bobo-Jimenez et al., 2017) that accumulates in the brain of AD patients (Herskowitz et al., 2013). To do so, we first treated neurons with Aβ25-35 and found that Rock2 protein levels time-dependently increased (Figure 3A) through a mechanism that could be prevented with roscovitine (Figure 3B). Moreover, inactivation of APC/C-Cdh1 with ProTAME promoted the accumulation of its cognate targets, Cdh1 and Rock2, an effect that was maintained during all the time of Aβ25-35 presence (Figure 3C). To ascertain whether Aβ25-35-induced Rock2 accumulation was translated into Rock2 activity, we measured the Thr853 phosphorylation of the Rock2-subtrate myosin phosphatase myosin-binding subunit (MBS) (Tanaka et al., 2006; Bobo-Jimenez et al., 2017). As shown in Figure 3D, Aβ25-35 increased MBS phosphorylation, indicating increased Rock2 activity, and effect that was potentiated by ProTAME. These results confirm that Aβ25-35 inactivates APC/C-Cdh1 to regulate Rock2 activity (Bobo-Jimenez et al., 2017). We then aimed to investigate whether the Cdh1-Rock2 axis impacts on Aβ25-35 neurotoxicity. To do this, neurons were treated with the clinically approved Rock inhibitor, fasudil, and with the Rock2 inhibitor, SR3677. As shown in Figures 4A,B, these treatments partially rescued Aβ25-35-induced neuronal apoptosis. Furthermore, either fasudil or SR3677 reduced Aβ25-35 neurotoxicity in Cdh1-knocked down neurons (Figures 4A,B). Neuronal viability was not affected by fasudil or SR3677 treatments (Supplementary Figure S2D). Although SR3677 demonstrated an approximately eightfold selectivity of Rock2 over Rock1, at a concentration of 10 µM it is likely that SR3677 also inhibits Rock1 (Feng et al., 2008). Next, we specifically knocked down Rock2 in neurons (siRock2), which did not affect Rock1 levels (Supplementary Figure S1C). Interestingly, Rock2 knock down also protected neurons against Aβ25-35-induced apoptotic death (Figures 4C,D), an effect that was mimicked by the APC/C-Cdh1 inhibitor ProTAME (Figure 4D). Altogether, these results indicate that Aβ25-35-induced Cdh1 phosphorylation inactivates APC/C-Cdh1, which results in Rock2 stabilization and activation, eventually leading to neuronal apoptosis.

Rock2 inhibition rescues cognitive impairment induced by Aβ in vivo. Finally, we asked whether the Aβ 25–35-Cdh1-Rock2 pathway herein described takes place in vivo. To this end, saline (control) or oligomerized Aβ25-35 were intracerebroventricularly injected in the mouse brain as previously done (Lapresa et al., 2019). One day after injection, we found that both Cdh1 and Rock2 were accumulated in the hippocampus (Figure 5A, left panel); however, at day 3, Cdh1 and Rock2 levels decreased to normal values (Figure 5, middle panel) and, at day 5, Cdh1 levels decreased below controls whereas Rock2 levels were maintained (Figure 5A, right panel). The increase in Rock2 abundance was paralleled by enhanced activity, as revealed by the increase in MBS phosphorylation in the hippocampus that was maintained after 5 days of injection (Figure 5B). To investigate whether Rock2 activation is implicated in the cognitive impairment caused by Aβ25-35, the Rock2 inhibitor SR3677 was intracerebroventricularly administered in the mice together with the Aβ25-35 oligomers. We found that Rock2 inhibition prevented the decrease in the novel object explorations and discrimination index caused by Aβ25-35 (Figure 6A), indicating an improvement in memory performance. In the Barnes maze test, Aβ25-35-injected mice displayed fewer time in scape (target) quadrant than control mice, indicating spatial memory deficits, an effect that was rescued by Rock2 inhibition (Figure 6B). No alterations in the time spent, distance run and number of entries in the center distance were found in the open field test, discarding any motor impairment in this model (Figure 6C). Together, these results indicate that Aβ25-35 enhances Rock2 activity in the mouse hippocampus causing memory loss.

Here, we characterize a novel signaling pathway involved in Aβ neurotoxicity that may open new therapeutic strategies to combat cognitive impairment in AD. We describe that Aβ25-35 oligomers induce cyclin dependent kinase-5 (Cdk5)-mediated phosphorylation of the APC/C-cofactor, Cdh1, leading to inhibition of APC/C and, eventually, neuronal apoptosis. Moreover, we show that Aβ25-35-induced APC/C-Cdh1 inhibition causes Rock2 accumulation and activation in neurons. Finally, we demonstrate that the memory loss caused by Aβ25-35 can be prevented by a pharmacological approach based on the selective inhibition of Rock2 activity. Given that Rock2 accumulates in the neurons of early-stage human AD brain (Herskowitz et al., 2013) and is associated with AD hallmarks (Gao et al., 2019; Cai et al., 2021), our data showing efficacy of the Rock2 inhibitor, compound SR3677, against Aβ25-35 memory impairment in mice should be considered as a potential therapeutic approach in AD.

We herein also characterize the signaling pathway that connects Aβ25-35 with Rock2. It was known that dysregulation of Ca²⁺ homeostasis is a key event in AD pathogenesis (LaFerla 2002) and that Aβ causes p35 cleavage to p25 -a potent Cdk5 activator- in a Ca²⁺-dependent manner (Fuchsberger et al., 2016; Lapresa et al., 2019). Notably, p25 accumulates, and Cdk5 is active, in the brain of AD patients before the onset of clinical symptoms (Patrick et al., 1999). In good agreement with these previous findings, here we demonstrate that the activation of Cdk5 by Aβ is a rapid process, as is the phosphorylation of Cdk5 substrate, Cdh1. Once phosphorylated, Cdh1 becomes disassembled from the APC/C complex, inhibiting its ubiquitin ligase activity for a long period of time. Given that Rock2 is a substrate of APC/C-Cdh1 complex, its inhibition by Cdh1 phosphorylation causes stabilization of Rock2 protein, which is maintained elevated and active also for long periods. Interestingly, this sequence of events mirrors that of Rock2 in AD patients, in which Rock2 is found elevated from the early stages of the disease and remains high throughout the AD progression (Herskowitz et al., 2013). This analogy between our data and those found in humans suggest that the early biochemical changes that we describe, namely Cdk5 activation and Cdh1 phosphorylation, might be considered as potential biomarkers of early AD detection and progression.

Our work also describes an intriguing biphasic effect of the temporal changes in Cdh1 protein abundance by Aβ25-35. Thus, Cdh1 is known to be a substrate of APC/C-Cdh1 and, therefore, when APC/C-Cdh1 is active, Cdh1 undergoes autoubiquitination and proteasomal degradation (Veas-Perez de Tudela et al., 2015b), which is considered a major regulatory system of Cdh1 homeostasis in neurons. Accordingly, in our hands, Aβ25-35 initially triggered APC/C-Cdh1 inactivation, which explains Cdh1 accumulation at the early time points after Aβ exposure, both in vitro and in vivo. However, at longer time periods, Cdh1 proteins levels underwent a progressive decrease, a result that is compatible with a previous study (Fuchsberger et al.) in which it was found that long-term incubation of neurons with Aβ oligomers causes a proteasome-dependent degradation of Cdh1. These observations suggest that Cdh1 may be subjected to regulation by alternative ubiquitin ligase(s) that could explain the maintained low protein levels of Cdh1. Notably, the F-box and WD-40 domain protein 11, FBXW11, has been observed in the hippocampus of AD mouse models (Sun et al., 2021). Whether FBXW11, which ubiquitinates phosphorylated substrates, including Cdh1 -in proliferative cells (Fukushima et al., 2013)- is responsible for the long-term neuronal degradation of Cdh1, is an interesting possibility that remains to be investigated in the context of AD pathogenesis.

Finally, our data may also provide clues for future development of pharmacological strategies aimed to specifically interfere in the Cdk5-Cdh1-Rock2 axis. Thus, Cdh1 needs to be phosphorylated by Cdk5 to be inactive and, therefore, to maintain Rock2 elevated upon Aβ treatment. These specific Cdk5-dependent phosphorylation sites on Cdh1 are Ser-40, Thr-121 and Ser-163 (Maestre et al., 2008). In fact, Aβ-induced neuronal apoptosis was prevented by expressing a triple phosphodefective (Ser⁴⁰Ala, Thr¹²¹Ala and Ser¹⁶³Ala) Cdh1 mutant, but not by expressing a triple phosphomimetic (Ser⁴⁰Asp, Thr¹²¹Asp and Ser¹⁶³Asp) Cdh1 mutant. We propose that these findings may be useful for the design of small molecules aimed to competitively and selectively antagonize these Cdh1 phosphorylation sites. If so, the subsequent disruption of the Cdk5-Cdh1 pathway might eventually serve to prevent aberrant Rock2 accumulation and neurodegeneration in AD.

In summary, our results provide evidence for a key role of the Cdh1-Rock2 signaling pathway in mediating neuronal apoptosis and memory impairment caused by Aβ25-35 oligomers. Thus, Aβ25-35 phosphorylates and inactivates Cdh1, which results in Rock2 stabilization and activation, leading to neurodegeneration. As Rock2 accumulates in the earliest stages of AD and remains elevated throughout the disease progression (Herskowitz et al., 2013), our data set the basis for a future development of therapeutic strategies to overcome neuronal loss and memory impairment in AD.