Current Alzheimer Research (www.bentham.org/car), 2004, 1, 33-38
Bentham Science Publishers Ltd.(www.bentham.org)

---

Recent Evidence Regarding a Role for Cdk5 Dysregulation in Alzheimer’s Disease

E.A. Monaco III^*

Department of Pharmacology, SUNY Upstate Medical University, Syracuse, NY 13210, USA

*Address correspondence to this author at the Department of Pharmacology SUNY Upstate Medical University Syracuse, NY 13210, USA; Tel: (315) 464-7970; Fax: (315) 464-8014; E-mail: monacoe@upstate.edu

Abstract: Based on a growing literature, cyclin-dependent kinase 5 (Cdk5) has been implicated in the pathological processes that contribute to neurodegeneration in Alzheimer’s disease (AD). Cdk5 is ubiquitously expressed, but its activity is largely localized to post-mitotic neurons due to neuron-specific expression of its activators p35 and p39. Sufficient Cdk5 activity is critical to normal central nervous system development, as in its absence, neuronal migration and axonal path finding are deranged. Conversely, excessive and mislocalized Cdk5 activity appears to be detrimental to neuronal function. In fact, the pathological hallmarks of AD, b-amyloid aggregates and neurofibrillary tangles, have been linked to Cdk5-mediated neuronal death. In this model, b-amyloid is the toxic stimulus that disrupts intracellular calcium homeostasis, leading to activation of calpains, a family of calcium-dependent proteases. Calpain-mediated cleavage of p35, yields a truncated p25 fragment that possesses a longer half-life, lacks the necessary sequence targeting it to membranes, but retains the capacity to activate Cdk5. The resulting excessive and mislocalized Cdk5 activity targets tau as a substrate for hyperphosphorylation, which is a prerequisite of paired helical filament (PHF) formation. A number of recent reports, utilizing diverse methods, lend further support to this model of AD neurodegeneration, and several strategies for combating Cdk5 dysregulation have even been devised. However, the study of Cdk5 in AD is not without controversy, and questions remain regarding its role in the pathology. Herein, the most recent findings regarding this model are reviewed.

Introduction

Cyclin-dependent kinase 5 (Cdk5) is an unique member of the cyclin-dependent kinase family of proline directed serine/threonine kinases. Unlike its family members, Cdk5 has no known cell-cycle functions, and instead is most highly expressed and active in post-mitotic neurons [1, 2, 3]. Moreover, where other Cdks require cyclins for enzymatic activation, the activators of Cdk5 are the non-cyclins p35 and p39 [4, 5]. Cdk5 is a multifunctional kinase, and phosphorylates an ever-growing list of substrates involved in cellular functions ranging from microtubule stability to N-methyl-D-aspartate receptor responsiveness (reviewed in [6, 7, 8]). With its diverse substrates and ubiquitous localization in neurons, it is not surprising that Cdk5 is critical for nervous system development. Null mice lacking Cdk5 demonstrate perinatal mortality associated with nervous system defects, including absent cortical laminar organization and cerebellar folliation, and abnormal neuronal migration [9]. Mice lacking p35 manifest a similar, albeit less severe phenotype, likely the result of functional redundancy with, or compensation by, p39 [10]. However, an emerging body of data supports a role for Cdk5 in neurodegenerative diseases like Parkinson’s, amyotrophic lateral sclerosis, and the subject of this review, Alzheimer’s disease (AD).

The Model

The first indication that Cdk5 might be involved in AD pathology came from the finding that it co-purifies with tau and recognizes it as a substrate in vitro [11]. Several of the tau sites phosphorylated by Cdk5 in vitro are the same sites phosphorylated in tau from paired helical filaments (PHFs) in AD tissue [12]. Although a number of kinases are able to phosphorylate AD-related tau epitopes (reviewed in [13]), Cdk5 is particularly relevant because its expression is specific to neurons. Interestingly, tau is unlikely to be a physiological substrate of Cdk5/p35. For example, co-transfection of COS-7 cells with a human tau construct and Cdk5/p35 yielded very little phosphorylation of the AT8 (serines 202 and 205) and PHF-1 (serines 396 and 404) tau phosphoepitopes [14]. Rather, Cdk5 efficiently phos-phorylates tau when complexed with p25, a proteolytic cleavage product of p35 [14]. To generate p25 from p35, the amino terminal domain is cleaved off by calpain, a calcium dependent cysteine protease [15, 16, 17]. While p25 is sufficient to activate Cdk5, it does not possess the N-terminal myristolation sequence that targets p35 to cellular membranes, and normally restricts access of Cdk5/p35 to substrates in this subcellular locale [14]. Thus p25 demonstrates a profoundly different subcellular distribution, localized to the cytosol rather than near membranes, allowing it access to a wider array of substrates. Another important difference between the two proteins is relative stability. Compared to p35, p25 has a half-life approximately 5 to 10 times greater [14]; the net result being that Cdk5 activity is sustained. Taken together, the data indicate that calpain-mediated cleavage of p35 yields excessive and mislocalized Cdk5 activity, potentially resulting in phosphorylation of a number of non-physiological substrates like tau.

Indeed, AD associated neuropathology is positively correlated with the presence of Cdk5/p25 complexes. Phosphorylation of tau on AD specific phosphoepitopes, both in transfected cells and primary neurons, has been reported [14]. Expression of Cdk5/p25 complexes induces morphological and cytoskeletal disruption, and ultimately apoptosis, in primary cortical neurons [14]. Moreover, neurotoxic insults that disrupt calcium homeostasis (i.e. maitotoxin, ischemia, and glutamate) activate calpain and generate p25 [15, 16, 17]. Utilizing a model that more closely recapitulates AD neuropathology, the application of fibrillogenic b-amyloid peptides to neuronal cultures, it has been shown that b-amyloid induces neurotoxicity accompanied by calpain activation, p35 cleavage to p25, and tau hyperphosphorylation [15, 18, 19].

Bolstering in vitro experimentation, Cdk5 dysregulation has been observed in transgenic mice and postmortem AD brain tissue. Expression of human p25 in transgenic mice resulted in increased tau phosphorylation at AD associated AT8 and PHF13 phosphoepitopes in amygdala, thalamus/hypothalamus, and cortex [20]. Importantly, this tau phosphorylation overlapped with increases in Cdk5 activity and areas of axonal swelling and neuronal cytoskeletal disorganization, suggesting that Cdk5/p25 complexes may play a role in AD tau hyperphosphorylation. Ultimately, if Cdk5 dysregulation is critical to AD neuropathology, evidence should be present in human postmortem brain tissue. Indeed, in seven of eight AD patients’ samples, there was a 20-40-fold increase in p25 compared to p35, relative to control samples [14]. Moreover, many neurons with p25 accumulation also contained neurofibrillary tangles [14]. Together, these findings support a model of AD pathology wherein b-amyloid induces Cdk5 dysregulation via calpain-mediated generation of p25, subsequent Cdk5/p25-mediated hyperphosphorylation of tau that renders it unable to stabilize microtubules, leading to neuronal demise (Fig. 1).

Since these first reports, a number of investigators have also found important links between aberrant Cdk5 activity and AD, making Cdk5 an increasingly attractive target for AD therapy.

Recent Evidence Supporting the Model

A major paradox in our understanding of this model of Cdk5 dysregulation is the observation that AD symptoms typically become manifest in the later decades of life. Indeed, this extended latency is one of the major obstacles for any theory involving AD neuropathology. In particular, if Cdk5 dysregulation is responsible for neurodegeneration, when does this process begin? This question is beginning to be understood. The levels of p35 protein required to activate Cdk5 are tightly regulated, primarily by limiting its availability. Cdk5 itself contributes to this process. It not only recognizes a variety of cellular substrates, it phosphorylates p35, which targets it for degradation by the ubiquitin-proteasome pathway [21]. This accounts for the relatively short half-life of p35 and provides a feedback mechanism for regulating Cdk5 activity. However, the phosphorylation state of p35, and thus its proteolysis, appear to be developmentally regulated [22]. In fetal rat brain, p35 is phosphorylated on distinct sites that promote proteasomal degradation and confer calpain resistance, respectively. In contrast, p35 in adult rat brain is unphosphorylated at either site. A potential consequence of this developmental shift is that p35 proteolysis shifts from the proteasome to calpains. Thus, maturation dependent changes in p35 phosphorylation state may be a mechanism by which adult brain becomes susceptible to Cdk5 dysregulation. However, of what physiological consequence this maturation-dependent phenomenon is, remains to be determined. Another question involves the other major Cdk5 brain-specific activator, p39. Does it contribute to Cdk5 dysregulation under conditions in which calpain proteolyzes p35?

The fact that p35 null mice are less severely affected than Cdk5 knockouts suggests the existence of functional redundancy, likely resulting from p39 [10]. Also, it appears that the functional redundancy of p35 and p39 can be extended to Cdk5 dysregulation in the AD model. For example, p39 proteolysis by calpain yields a c-terminal fragment, p29, which becomes mislocalized and is more stable [23]. In vitro, three neurotoxic insults that generate p25 (ionomycin, glutamate, and ischemia [15]), also induce p29 [23]. Thus, Cdk5 dysregulation could result from calpain-mediated proteolysis of either of its major activators, p35 or p39.

It is important to note that neuronal pathology in the aged brain likely precedes the formation of senile plaques composed of b-amyloid [24, 25, 26, 27]. As such, Town et al. [28] examined the effects of soluble b-amyloid on Cdk5 activity in p35 over expressing N2a cells, recapitulating a condition prior to b-amyloid deposition. Following application of soluble b-amyloid, the Cdk5/p35 axis was perturbed, yielding increases in the p25/p35 ratio, Cdk5 activity, and tau hyperphosphorylation. Consistent with the involvement of calcium homeostasis, the effects of soluble b-amyloid on Cdk5 signaling were significantly attenuated by the calcium channel blocker, verapamil, and calpain inhibitor I. These findings indicate that b-amyloid mediates Cdk5 dysregulation via calpain activation early in the development of AD before senile deposits are detectable.

Mislocalization and sustained activity appear to be the modes by which Cdk5 become dysregulated by p25. However, Hashiguchi et al. [29] provided evidence that the kinetics of Cdk5 catalysis are also affected by p25. Utilizing an in vitro phosphorylation assay with human tau as a substrate, they show that Cdk5/p25 complexes phosphorylate tau at a faster rate than Cdk5/p35 complexes (p25_kcat/km>p35_kcat/km). Moreover, after saturation, the stoichiometry of Cdk5/p25 phosphorylation is greater than that for Cdk5/p35 (3.3 vs. 2.3 moles per 1 mol tau). Finally, they suggest that unlike p35, p25 promotes sequential phosphorylations of the AD phosphoepitopes serine 202 and threonine 205. The result is a kinase that is not only mislocalized and possessed of sustained activity, but also one that shows enhanced activity for tau. While in vitro evidence is mounting, it is nonetheless critical to follow-up with similar studies in more complex animal models.

Although tau abnormalities have been observed in transgenic mice with increased Cdk5 activity, neurofibrillary tangle formation was not observed [20]. For example, Bian et al. [30], reported focal accumulations of tau and axonal degeneration in p25 transgenic mice, but found no evidence of neurofibrillary tangles, despite five fold overexpression of the protein. Such reports suggest that while Cdk5 dysregulation may contribute to tau hyperphosphorylation, it is, by itself, insufficient to generate neurofibrillary tangles. Extending these findings, Noble et al. [31] crossed p25 overexpressing transgenic mice with those overexpressing human tau (P301L). Using this background, they observed tau hyperphosphorylation, as well as an increase in the formation of neurofibrillary tangles, versus single transgenics. Thus, although Cdk5 alone may be insufficient to induce neurofibrillary tangles, its activity can increase their formation. Interestingly, Noble et al. [31] also point to a sequential mechanism in which phosphorylation by Cdk5, in concert with other tau kinases like glycogen synthase kinase 3b (GSK3b), induces neurofibrillary tangle formation. This is consistent with much evidence linking GSK3b to tau hyperphosphorylation and AD neuropathology (reviewed in [32]).

Transgenic mice overexpressing the Swedish familial AD double mutation in the human b-amyloid precursor protein progressively develop typical b-amyloid deposits, associated with age-related cognitive impairment [33, 34, 35]. Analysis of changes in the Cdk5/p35 axis in these mice demonstrated that p25 accumulated, versus wild-type animals, without a change in Cdk5 protein levels [36]. Moreover, the transgenic animals showed increased Cdk5 activity and tau phosphorylation at the AT8 and PHF-1 phosphoepitopes, findings supporting a relationship between b-amyloid and Cdk5 dysregulation. Despite the emerging evidence linking Cdk5 dysregulation to AD neuropathology, some recent reports are confounding, especially analyses of postmortem human brain tissue.

Difficulties Regarding the Model

Neurotoxic insults that disrupt calcium homeostasis (i.e. glutamate and ionomycin) induce p25 generation by calpain, and subsequent tau hyperphosphorylation [15, 16, 17]. However, Kerokoski et al. [37] found that although treatment of hippocampal cultures with glutamate, N-methyl-D-aspartate, or ionomycin resulted in p25 formation and increased Cdk5 activity, tau hyperphosphorylation was not apparent. A previous study in which rat hippocampal neurons were exposed to the calcium ionophore A23187 yielded similar results [38]. Thus, while not precluding a role for Cdk5, these results suggest that tau hyperphosphorylation may not invariably occur as a result of p25 generation. They also indicate that in vitro neurotoxicity models wherein calpain is activated may not sufficiently mimic AD phenomena.

As mentioned, transgenic mouse models of AD have yielded supportive, yet at times confounding results (e.g. p25 transgenics do not display neurofibrillary tangles). However, Takashima et al. [38], utilizing a different p25 overexpressing transgenic mouse, not only failed to observe neurofibrillary tangle formation, but they were unable to detect any neuropathology, even in the face of 3-5 fold overexpression of p25 and increased Cdk5 activity. In another study, utilizing a mouse model similar to that of Otth et al. [36], overexpressing a triple mutant form of the b-amyloid precursor protein, Tandon et al. [39] did not detect any changes in p35, p25, or Cdk5. Reports such as these point to the complexity of transgenic mouse models and encourage an understanding of the differences that contribute to divergent results. For example, the mouse used by Ahlijanian et al. [20] was developed from a human p25 gene fragment, whereas Takashima et al. [38] used a bovine p25 cDNA.

Postmortem Human Brain Studies

Arguably the most perplexing of the studies regarding Cdk5 dysregulation in AD have been those in which postmortem human brain tissue has been analyzed. This is best illustrated by a timeline of the findings (Fig. 2). The first such study, by Patrick et al. [14], surveyed eight samples from patients with AD, four from control individuals, and one from a patient with Huntington’s disease (HD) (samples were taken from Brodmann areas 11, 21, and 45). p25 was found to accumulate 20-40 fold, compared to p35, in all but one patient with AD. In contrast, the four controls and one HD sample showed no detectable changes.

In light of calpain’s role in Cdk5 dysregulation [15, 17], Taniguchi et al. [40] investigated p35 cleavage and calpain activation in control and AD brain. The eight AD samples examined showed no increase in the ratio of p25 to p35 versus the nine age-matched controls. Instead, a 50% decrease in the total levels of p25 and p35 was observed in the AD samples, in contrast to the seven-fold increase in activated m-calpain. Independently, Yoo and Lubec [41], in their study of samples from seven Down’s syndrome (DS) patients (for a review on the DS model of AD see [44]), six AD patients, and eight age matched controls, also observed a reduction in p25 levels in both DS and AD patients. Takashima et al. [38], in a comparison of parietal cortex from five AD brain samples and five controls, also failed to demonstrate an detectable accumulation of p25 in AD tissue. Nguyen et al. [42] examined the prefrontal cortex, an area heavily involved in the degeneration of AD, and cerebellum, spared in AD, and compared changes in p25 levels in 10 AD and 10 control subject tissues. Again, no significant changes in p25 levels were detectable.

To address these divergent findings, Tseng et al. [43] performed a larger study on samples from 28 AD patients and 25 age-matched controls. They observed that mean p25/p35 ratios were higher in AD versus controls. Moreover, the p25/p35 ratios were consistently higher in AD versus controls across all three brain regions examined (frontal cortex, inferior parietal cortex, and hippocampus), thus extending their previous findings [14]. However, the most recent report delivers yet another confounding result. Tandon et al. [39], in a comparison of p25 levels in several neurodegenerative tauopathies, including AD, observed that the p25/p35 ratio was not significantly different between the tauopathies and controls. Analysis of AD subtypes (13 familial versus 11 sporadic AD) also failed to demonstrate a significant difference in the p25/p35 ratio between AD and controls.

In the end, we are left to ponder the source of these differences. It is likely that post-mortem intervals (PMIs) affect measurements of the p25/p35 ratio. Patrick et al. [45] suggest that in samples with long PMIs (i.e. 24-hrs), observable changes in the p25/p35 ratio between AD and control samples become less marked, and can disappear altogether. After as little as 1-hr after death, and peaking after 14-hrs, p25 is generated from p35 [40]. This p35 degradation is caused by artifactual postmortem calpain activity, as it was significantly attenuated by methods to inhibit these proteases. However, despite short PMIs in their study, Taniguchi et al. [40] found no increases in p25 in AD brain (Table 1). These results indicate that even short PMIs can result in changes to tissue biochemistry that make p25 comparisons extraordinarily difficult.

The extent of AD progression could also confound analyses of postmortem brain. The bulk of the evidence indicates that the biochemical events responsible for AD occur early in the disease. Consistent with this, Patrick et al. [14] found that tissue from a patient with late-stage AD, and extensive neuronal loss, did not show an increase in p25 like the other AD samples studied. In this case, they concluded that p25 likely contributes early in AD, and is not demonstrable after extensive neuronal loss has already occurred. As such, in designing similar studies, it would be appropriate to evaluate the extent of AD neuropathology, limiting analysis to samples of younger patients with less severe disease.

In addition, it is necessary for researchers in this field to utilize available methods for controlling potential confounds in postmort