p35/p25 Is Not Essential for Tau and Cytoskeletal Pathology or Neuronal Loss in Niemann–Pick Type C Disease

Abstract

Hyperactivation of the cyclin-dependent kinase 5 (cdk5), triggered by proteolytic conversion of its neuronal activator, p35, to a more potent byproduct, p25, has been implicated in Alzheimer's disease (AD), amyotrophic lateral sclerosis, and Niemann–Pick type C disease (NPC). This mechanism is thought to lead to the development of neuropathological hallmarks, i.e., hyperphosphorylated cytoskeletal proteins, neuronal inclusions, and neurodegeneration, that are common to all three diseases. This pathological ensemble is recapitulated in a single model, the npc-1 (npc^−/−) mutant mouse. Previously, we showed that pharmacological cdk inhibitors dramatically reduced hyperphosphorylation, lesion formation, and locomotor defects in npc^−/− mice, suggesting that cdk activity is required for NPC pathogenesis. Here, we used genetic ablation of the p35 gene to examine the specific involvement of p35, p25, and hence cdk5 activation in NPC neuropathogenesis. We found that lack of p35/p25 does not slow the onset or progression or improve the neuropathology of NPC. Our results provide direct evidence that p35/p25-mediated cdk5 deregulation is not essential for NPC pathology and suggest that similar pathology in AD may also be cdk5 independent.

Introduction

Abnormal aggregates of neuronal cytoskeletal proteins are signatures of many human neurodegenerative diseases. Neurofibrillary tangles (NFTs) composed of hyperphosphorylated tau are observed in Alzheimer's disease (AD) and tauopathies (Lee et al., 2001). Axonal spheroids containing neurofilaments (NFs) and hyperphosphorylated tau are observed in motor neuron diseases such as amyotrophic lateral sclerosis (ALS) (Delisle and Carpenter, 1984) and tauopathies (Axelsson et al., 1984; Zhou et al., 1998). Curiously both NFTs and spheroids are produced concomitantly in Niemann–Pick type C disease (NPC), a fatal neurovisceral lipidosis with widespread neurodegeneration. Antigenically and ultrastructurally, the NFTs of NPC are similar to those of AD and tauopathies, and the spheroids of NPC resemble those of ALS and tauopathies (Bu et al., 2002b; Walkley and Suzuki, 2004). Thus, NFTs and spheroids may be produced by similar posttranslational mechanisms in these etiologically diverse conditions.

Cyclin-dependent kinase 5 (cdk5) is thought to be responsible for tau hyperphosphorylation in these diseases (Cruz and Tsai, 2004). Cdk5 is activated by p35 or p39, both of which may be cleaved to more stable and potent fragments, p25 and p29 (Cruz and Tsai, 2004). Elevated p25 and cdk5 hyperactivation are seen in AD (Patrick et al., 1999), ALS (Nguyen et al., 2001), and NPC (Bu et al., 2002a). Efforts aimed at supporting a causal relationship between cdk5/p25 and NFTs have relied on overexpression of cdk5, p35, or p25. Although some of these models have yielded positive results (Cruz et al., 2003; Noble et al., 2003), others have not (Ahlijanian et al., 2000; Takashima et al., 2001; Bian et al., 2002).

We tackled this relationship by using an alternative approach of inhibiting cdk5 activity in a natural model replicating neurodegenerative features of human disease, the npc^−/− mouse. This model arose because of spontaneous mutations in the npc1 gene, the orthologous gene mutated in human NPC. The NPC1 locus encodes a late endosomal/lysosomal lipid transporter (Walkley and Suzuki, 2004), and NPC1 mutations lead to lipid accumulation in all body tissues, weight loss, mental retardation or dementia, and motor defects with ataxia. npc^−/− mice recapitulate most features of human NPC. Although they do not develop the ultrastructural characteristics of human NFT (German et al., 2001), they do accumulate hyperphosphorylated tau in neuronal somata and form axonal spheroids. Tau from npc^−/− mouse brain appears similar to AD or human NPC tau by multiple qualitative and quantitative criteria (Bu et al., 2002a,b;). These similarities suggest that signaling mechanisms leading to tau modification in npc^−/− mice may be similar to those in the human diseases.

In vivo approaches are necessary for unraveling neurodegenerative mechanisms in NPC, because primary neurons from npc^−/− mice fail to develop cytoskeletal pathology and survive normally in vitro (Henderson et al., 2000). We previously assessed the role of cdk activity in NPC by intracerebroventricular treatment of npc^−/− mice with the pan-cdk inhibitors roscovitine or olomoucine (Zhang et al., 2004). These inhibitors attenuated cytoskeletal pathology and neurodegeneration, indicating an essential role for cdks in NPC.

Toward defining which cdk is involved and to determine specifically whether p25 is required, we crossed npc^−/− mice with p35 null mice. Because p25 is a derived proteolytically from p35, p35 null mice also lack p25. Constitutive cdk5 depletion is lethal (Ohshima et al., 1996), but p35 null mice are viable (Hallows et al., 2003). Behavioral, biochemical, and immunohistochemical comparisons of wild-type (wt), p35^−/−, npc^−/−, and npc^−/−, p35^−/− littermates show that lack of p35/p25 does not slow the development of NPC or inhibit tau phosphorylation or lesion formation.

Materials and Methods

Mice.

Heterozygous BALB/cNctr-Npc1^mIN/J (npc^+/−) mice were obtained from The Jackson Laboratory (Bar Harbor, ME) and maintained in our colony. npc^+/− mice were crossed with p35^+/− mice having a C57BL/6 background (Hallows et al., 2003) to obtain double heterozygous (npc^+/−, p35^+/−) mice in the mixed BALB/c/C57BL/6 background. Double-heterozygous mice were mated to obtain wt, p35^−/−, npc^−/−, and npc^−/−, p35^−/− genotypes in the mixed background, and all comparisons between these different genotypes were made using littermates. It should be noted that the mixed background did not have an observable effect on the basic phenotypic features of either npc^−/− or p35^−/− mice. A more direct test of the effects of the C57BL/6 background on the phenotype has been done by backcrossing the BALB/c line into the C57BL/6 strain (Walkley and Suzuki, 2004). The disease phenotype was similar except for an increase in spheroids in the fimbria. The comparisons made in the present study did not involve this region of brain. Other transgenic or knock-out mice in the C57BL/6 background have also been crossbred with the npc^−/− BALB/c mice, without any noticeable effect on NPC pathology (Erickson and Bernard, 2002). All protocols and housing were approved by the Animal Care and Use Committee at the University of Washington.

Genotyping was performed by PCR analysis of genomic DNA isolated from tail biopsies obtained at the time of weaning as reported previously [npc (Loftus et al., 1997); p35 (Hallows et al., 2003)].

Eight- to 9-week-old male and female mice were killed by decapitation, and brains were quickly dissected and halved sagittally. One-half was flash frozen for biochemical analyses, and the other half were immersion fixed overnight in 4% paraformaldehyde at 4°C, paraffin embedded, cut in 10 μm sections, and mounted on slides for histological analyses.

Antibodies.

Table 1lists all antibodies used for Western blotting. The NF SMI32 and the ganglioside GM2 antibodies used for histology were from Sternberger Monoclonals (Lutherville, MD) and Matreya (State College, PA), respectively. Subtype-specific secondary antibodies conjugated to HRP or biotin and streptavidin–HRP were from Southern Biotechnology (Birmingham, AL).

Table 1.

Antibodies

Group	Antibody	Isotype	Epitope	Dilution	Source
cdk5	cdk5 (DC 17)	Mouse IgG1	Full-length cdk5	1:500	Santa Cruz Biotechnology (Santa Cruz, CA)
	cdk5 (C-8)	Rabbit polyclonal	cdk5 C terminus	1:100	Santa Cruz Biotechnology
p35/p25	p35 (C-19)	Rabbit polyclonal	p35 C terminus	1:100	Santa Cruz Biotechnology
	p25N	Rabbit polyclonal	p25 N terminus	1:250	K. Ishiguro
p39	p39	Rabbit polyclonal	p39 N terminus	1:1000	K.-Y. Lee
Tau	TG5	Mouse IgG1	Total tau	1:10	I. Vincent
	CP13	Mouse IgG1	Phospho-tau Ser-202	1:10	P. Davies
	PHF-1	Mouse IgG1	Phospho-tau Ser-396/404	1:20	P. Davies
	MC6	Mouse IgG1	Phospho-tau Ser-235	1:10	P. Davies
	CP10	Mouse IgM	Phospho-tau Thr-231	1:10	P. Davies
	CP22	Mouse IgM	Phospho-tau Thr-175	1:10	P. Davies
	PG5	Mouse IgG3	Phospho-tau Ser-409	1:5	P. Davies
Neuronal antigens	NeuN	Mouse IgG1	Neuronal nuclei	1:230	Chemicon (Temecula, CA)

Western blotting.

Western blotting was performed as reported previously (Hallows et al., 2003). Briefly, frozen brain halves were homogenized in buffer, and aliquots were stored at −70°C. Supernatants of brain homogenates were electrophoresed, transferred to nitrocellulose, and probed with the antibodies listed in Table 1. In all experiments, protein levels were normalized to the levels of the neuronal antigen neuronal-specific nuclear protein (NeuN). Two to three blots were run for each set of samples for each epitope examined (n = 6 wt, 6 npc^−/−, and 6 npc^−/−, p35^−/−).

In vitro kinase assays. Cdk5 was immunoprecipitated from 100 μg of brain homogenates using 3 μg of polyclonal antibodies (Table 1), and activity was assayed using histone H1 (Calbiochem, La Jolla, CA) as a substrate as reported previously (Hallows et al., 2003) (n = 4 wt, 4 p35^−/−, 4 npc^−/−, and 4 npc^−/− , p35^−/−).

Histology.

Sagittal brain sections were sequentially incubated with the phospho-tau antibodies CP13 or paired helical filament-1 (PHF-1) at 1:10, the NF antibody SMI32 at 1:500, or the GM2 antibody at 1:100, biotin-conjugated secondary antibodies, streptavidin–HRP, and then DAB to visualize antibody binding, followed by hematoxylin and eosin (H&E) counterstaining as reported previously (Hallows et al., 2003). For phospho-tau staining, sections were first incubated with 50 mm NH₄Cl for 30 min before incubation with the primary antibody. Images were collected using an Olympus Optical (Thornwood, NY) DP10 digital camera attached to an Olympus Optical BX40 microscope (n = 3 wt, 3 p35^−/−, 6 npc^−/−, and 6 npc^−/−, p35^−/−).

Purkinje neuron and axonal spheroid quantitation.

Sagittal brain sections 150 μm lateral to midline from 8- to 9-week-old mice were incubated with the SMI32 antibody to NFs to visualize Purkinje neurons and axonal spheroids. The numbers of surviving Purkinje neurons were counted in the third cerebellar lobe from two adjacent sections for each animal and averaged, the length of the margin between the molecular and granule cell layers was measured, and the number of Purkinje neurons per millimeter was calculated. The number of surviving Purkinje neurons per millimeter was averaged within genotypes (n = 8 npc^−/− and 8 npc^−/−, p35^−/−). For axonal spheroid quantitation, the numbers of axonal spheroids 8 μm or larger in a 925 × 740 μm field were counted in anatomically matched sections of three brain regions: central pons, cerebellar peduncles, and inferior colliculus. For each animal, spheroids were counted in two adjacent sections and averaged. The average number of spheroids per animal was then averaged within genotypes (n = 6 npc^−/− and 6 npc^−/−, p35^−/−).

Motor ability and weight loss.

Mice were weighed and motor ability was assessed twice weekly between 3 and 10 weeks of age. To assess motor ability, mice were allowed to grasp the bottom of a coat hanger suspended 18 cm above a flat surface and allowed to remain on the hanger for up to 2 min (hanging time). Their behavior on the hanger was also noted (see Fig. 4). Hanging time was multiplied by behavior to obtain a numeric value for motor ability. Values were averaged within genotypes at each time point (n = 13 wt, 14 p35^−/−, 10 npc^−/−, and 13 npc^−/−, p35^−/−).

Figure 4.

Weight and motor ability of wt, p35^−/−, npc^−/−, and npc^−/−, p35^−/− mice between 3 and 10 weeks of age. A, npc^−/− and npc^−/−, p35^−/− mice had similar weight loss. B, Motor ability deteriorated in npc^−/− mice beginning at ∼6 weeks of age, whereas npc^−/−, p35^−/− mice exhibited a decline in motor ability beginning at ∼5 weeks of age.

Statistical analyses.

All statistical analyses were performed using the Student's t test. Significance is indicated at *p < 0.05 or **p < 0.001. Sampling errors are reported as the SEM.

Results

Cdk5 activity is reduced in npc^−/−, p35^−/− mice

Cdk5 protein levels and activity were examined in supernatants from brain homogenates of npc^−/− and npc^−/−, p35^−/− mice (Fig. 1). Western blotting detected similar levels of cdk5 protein in brain supernatants of both genotypes of mice using two different cdk5 antibodies (Fig. 1A, shown for DC17). In contrast to npc^−/− mice, p35 and p25 were absent from npc^−/−, p35^−/− mice (shown for p25N). Levels of the cdk5 activator p39 were similar in both groups, ruling out compensatory upregulation attributable to lack of p35/p25.

Figure 1.

cdk5 levels and activity. A, Western blots showing similar levels of cdk5 and p39 in npc^−/− and npc^−/−, p35^−/− mouse brains. p35 and p25 were detected in npc^−/− mice but not in npc^−/−, p35^−/− mice. NeuN is shown as a protein loading control. B, Autoradiograph from an in vitro cdk5 kinase assay and graphic representation of the average cdk5 activity for each genotype. Cdk5 activity was reduced ∼78% in npc^−/−, p35^−/− mice compared with npc^−/− mice.

In vitro cdk5 kinase assays detected a 78.2 ± 8.9% reduction (p < 0.05) in cdk5 activity in npc^−/−, p35^−/− mice compared with npc^−/− mice (Fig. 1B), indicating that much of the cdk5 activity in npc^−/− mice is mediated by p35 or p25. The residual cdk5 activity in npc^−/−, p35^−/− mice was similar to that of p35^−/− mice and 32.5 ± 13.3% (p < 0.05) that of wt mice. The marked inhibition of cdk5 in npc^−/−, p35^−/− mice makes this model suitable for studying the role of p35/p25 and cdk5 in NPC.

Tau phosphorylation is not reduced in npc^−/−, p35^−/− mice

Previous studies found increased tau phosphorylation at several different epitopes in the brains of npc^−/− mice compared with wt (Bu et al., 2002a), and administration of cdk inhibitors in npc^−/− mice reduced phosphorylation at these epitopes in a dose-dependent manner (Zhang et al., 2004). If p35/p25-mediated cdk5 activity is responsible for hyperphosphorylation in NPC, npc^−/− mice with greatly reduced cdk5 activity should have reduced phosphorylation.

We examined levels of tau and tau phosphoepitopes in wt, npc^−/−, and npc^−/−, p35^−/− mouse brains (Fig. 2). We found similar levels of total tau (TG5) in wt and npc^−/− mice, but levels were increased in npc^−/−, p35^−/− mice, consistent with previous findings in p35^−/− mice (Hallows et al., 2003). After normalizing to TG5 immunoreactivity, tau phosphorylation in npc^−/− mice was significantly increased at Ser-202 (CP13) and Ser-396/404 (PHF-1) compared with wt. In npc^−/−, p35^−/− mice, tau phosphorylation at Ser-202 (CP13) was even further increased, almost threefold higher than wt and twofold higher than npc^−/− mice. Tau phosphorylation at Ser-396/404 was increased almost twofold in npc^−/−, p35^−/− mice compared with wt. These data demonstrate that hyperphosphorylation in npc^−/− mice occurs independent of p35/p25-mediated cdk5 activation.

Figure 2.

Tau phosphorylation. A, Western blots showing similar levels of total tau (TG5) in wt and npc^−/− mice but increased total tau (107.4 ± 40.4%; p < 0.05) in npc^−/−, p35^−/− mice compared with wt and npc^−/− mice. Tau phosphorylation at Ser-202 (CP13) was increased in npc^−/− mice (106.7 ± 37.6%; p < 0.05) and further increased in npc^−/−, p35^−/− mice (292.5 ± 55.9%; p < 0.001) compared with wt or increased 90.4 ± 27.1% (p < 0.05) compared with npc^−/− mice. Tau phosphorylation at Ser-396/404 (PHF-1) was increased in both npc^−/− (68.5 ± 31.7%; p < 0.05) and npc^−/−, p35^−/− (97.1 ± 19.8%; p < 0.001) mice compared with wt but was not significantly different between npc^−/− and npc^−/−, p35^−/− mice. Tau phosphorylation increased in both npc^−/− and npc^−/−, p35^−/− mice despite similar levels of cdk5 in all three genotypes of mice. NeuN is shown as a protein loading control. B, Sagittal brain sections stained with the tau phospho-Ser-202 antibody CP13 in wt, npc^−/−, and npc^−/−, p35^−/− brains in the pons (left; scale bar, 50 μm) and cortex (right; scale bar, 100 μm). CP13 immunoreactivity was seen in axonal spheroids (arrows) in the pons in both npc^−/− and npc^−/−, p35^−/− brains. Staining was more robust in npc^−/− brains compared with wt and was even more robust in npc^−/−, p35^−/− brains compared with npc^−/− brains. The disruption of the normal layered cytoarchitecture of the cortex seen in the npc^−/−, p35^−/− brain is a component of the p35^−/− phenotype (Hallows et al., 2003).

At the immunohistochemical level, CP13 and PHF-1 immunoreactivities were increased in npc^−/− mice compared with wt mice and were localized to axons and axonal spheroids (Fig. 2B, shown for CP13 only). In npc^−/−, p35^−/− mice, CP13 immunoreactivity was qualitatively similar but more intense than in npc^−/− mice in regions with cytoskeletal pathology (pons, left panels) and in the neuropil of regions without overt cytoskeletal pathology, i.e., the cortex (right panels). PHF-1 did not reveal any striking difference in npc^−/−, p35^−/− mice compared with npc^−/− mice (data not shown).

Other tau phosphoepitopes, i.e., Thr-175 (CP22), Thr-231 (CP10), Ser-235 (MC6), and Ser-409 (PG5), were also examined by Western blotting and immunohistochemistry. With both techniques, no significant changes were observed with these phosphoepitopes when npc^−/−, p35^−/− mice and npc^−/− mice were compared (data not shown).

Collectively, these data demonstrate that tau hyperphosphorylation in npc^−/− mice proceeds even in the presence of drastically reduced cdk5 activity.

Neuropathological hallmarks are not reduced in npc^−/−, p35^−/− mice

The brains of npc^−/− mice are characterized by the formation of several neuropathological lesions, including lipid accumulation within neurons, formation of axonal spheroids, and neuron loss, most prominently of cerebellar Purkinje cells (Walkley and Suzuki, 2004). Continuous intracerebroventricular administration of cdk inhibitors in npc^−/− mice resulted in improved Purkinje neuron survival and decreased axonal spheroid formation (Zhang et al., 2004). If p25-mediated cdk5 activity is involved in this pathogenesis, it is expected that formation of these pathologies would be reduced in npc^−/− mice lacking p35/p25.

The Purkinje neuron monolayer is normally situated between the molecular and granule cerebellar layers (Fig. 3A, wt). In p35 null mice Purkinje neurons are often misplaced throughout the granular layer and sometimes aggregate with a multilayered appearance, and granule cells are misplaced in the molecular layer (Hallows et al., 2003). This abnormal distribution of Purkinje cells is not seen in npc^−/− mice but is observed in npc^−/−, p35^−/− mice (Fig. 3A) and thus appears to be related to the absence of p35. Therefore, SMI32, an NF antibody that stains Purkinje cells (Gotow and Tanaka, 1994), was used to identify and count Purkinje cells in npc^−/− and npc^−/−, p35^−/− mice. We found a similar number of Purkinje neurons remaining in npc^−/− and npc^−/−, p35^−/− cerebellae, demonstrating that the lack of p35/p25 does not improve Purkinje neuron survival in npc^−/− mice.

Figure 3.

Pathological lesions. A, Sagittal brain sections stained with the NF antibody SMI32 showing normally (wt) and ectopically (p35^−/−) localized cerebellar Purkinje neurons (arrows). Many Purkinje neurons have been lost in npc^−/− and npc^−/−, p35^−/− mice, but labeled basket cell terminals still surround their former location (arrowheads). The ectopically localized granule cells in the molecular layer in the npc^−/−, p35^−/− is a component of the p35^−/− phenotype (Hallows et al., 2003). Scale bar, 50 μm. B, Sagittal brain sections stained with the NF antibody SMI32 showing no axonal spheroids in wt or p35^−/− mouse brains but numerous spheroids (arrows) in the pons of npc^−/− and npc^−/−, p35^−/− mice. Scale bar, 100 μm. C, Graph showing similar numbers of spheroids in the cerebellum (cb) and pons of npc^−/− and npc^−/−, p35^−/− mice. Spheroid numbers increased ∼73% in the inferior colliculus (ic) in npc^−/−, p35^−/− mice compared with npc^−/− mice. D, Sagittal brain sections stained with H&E showing similar lipid accumulation and formation of distended neurons in npc^−/− and npc^−/−, p35^−/− mice (arrows). Scale bar, 25 μm. E, Sagittal brain sections stained with an antibody to ganglioside GM2 showing similar accumulation of GM2 in npc^−/− and npc^−/−, p35^−/− mice (arrows). Scale bar, 50 μm.

Although both tau and NFs accumulate in spheroids, we found that SMI32 immunoreactivity (Fig. 3B) is a more reliable marker for detecting these lesions (Bu et al., 2002a). This antibody recognizes a nonphosphorylated sequence in the heavy subunit of NF, which is not affected by cdk inhibitor treatment (Zhang et al., 2004). SMI32-positive spheroids were not detected in wt or p35^−/− mice but were present in npc^−/− and npc^−/−, p35^−/− mice. Similar numbers were counted in the pons and cerebellum, but the inferior colliculus of npc^−/−, p35^−/− mice had a 72.8 ± 14.7% (p < 0.05) increase over npc^−/− mice (Fig. 3C).

NPC is primarily a lipid storage disease and is characterized by the formation of distended neurons containing nonesterified lipids and the ganglioside GM2 (Walkley and Suzuki, 2004). Although there is no reason presently to expect any change in lipid pathology by altering the function of cdk5, we examined lipid storage in npc^−/− and npc^−/−, p35^−/− mouse brains by H&E staining and GM2 immunostaining. Procedures involved in processing paraffin-embedded tissue usually lead to significant extraction of lipids, but evidence for lipid storage is seen by the foamy appearance of lipid in distended neurons and its negative reaction to H&E. These indices were similar in both groups of mice (Fig. 3D), as was GM2 immunoreactivity (Fig. 3E).

Weight loss and motor defects are not reduced in npc^−/−, p35^−/− mice

npc^−/− mice characteristically experience weight loss and a decline in motor ability beginning at ∼6 weeks of age (Loftus et al., 2002), and intracerebroventricular administration of cdk inhibitors in npc^−/− mice has been shown to improve both of these disease parameters (Zhang et al., 2004).

To determine whether weight loss or motor defects were improved in npc^−/− mice lacking p35/p25, we weighed mice and tested their motor ability twice weekly between the ages of 3 and 10 weeks of age. Both wt and p35^−/− mice continued to gain weight over the course of testing, whereas npc^−/− and npc^−/−, p35^−/− mice lost weight beginning at ∼6 weeks of age (Fig. 4A). Motor ability held fairly constant in wt and p35^−/− mice. npc^−/− mice demonstrated a dramatic decline in motor ability beginning at ∼6 weeks of age, whereas npc^−/−, p35^−/− mice demonstrated this decline at an earlier time of ∼5 weeks (Fig. 4B). Overall, these findings show no improvement in weight loss or motor ability in npc^−/− mice lacking p35/p25.

Discussion

Our current study demonstrates that the cardinal features of NPC, i.e., tau hyperphosphorylation, axonal spheroid formation, Purkinje neuron loss, weight loss, and motor defects, are not attenuated in npc^−/−, p35^−/− mice. Loss of p35/p25 not only negated the disease-associated elevation in cdk5 activity but reduced it to 32% of wt levels. The persistence of NPC pathology in the absence of p35/p25 proves unequivocally that p35/p25-mediated cdk5 activation is not responsible for hyperphosphorylation and subsequent cytoskeletal pathology in NPC.

Ruling out cdk5 entirely is more difficult. The 78% decrease in cdk5 activity in npc^−/−, p35^−/− mice, without an accompanying decrease in tau phosphorylation and axonal spheroid formation, suggests that cdk5 is not responsible for NPC pathology. Interestingly, phosphoepitopes thought to be produced by cdk5 [Ser-231 and Ser-235 (Lew et al., 1994)] and the proline-directed Thr-175 site were not affected; the CP13 Ser-202 site that can be produced by multiple proline-directed kinases was increased, and the PHF-1 epitope thought to be produced by glycogen synthase kinase 3β (GSK3β) (Lucas et al., 2001) was also increased. It could be argued that residual p35-independent cdk5 activity in npc^−/−, p35^−/− mice may be sufficient to drive pathology. We have shown (Hallows et al., 2003) that loss of p35 in a wt background redistributes cdk5 to more distal neuronal processes. This redistribution might concentrate p35-independent cdk5 activity to focal sites relevant to disease and improve its access to relevant substrates such as tau. In this scenario, the real culprit would be a lack or decreased levels of p35 rather than p25 accumulation or change in cdk5 activity. With respect to the latter, we (Hallows et al., 2003) and others (Sharma et al., 2002; Morfini et al., 2004) have proposed that cdk5 negatively regulates downstream kinases that phosphorylate tau. According to this scheme, a decrease in p35 or reduced cdk5 activity would promote cytoskeletal pathology and neurodegeneration. It is likely that the kinases activated by cdk5 inhibition may be responsible for the increased tau hyperphosphorylation at the CP13 and PHF-1 epitopes in npc^−/−, p35^−/− mice compared with npc^−/− mice. Because mice lacking cdk5 activity die at birth (Ohshima et al., 1996), a model of postdevelopmental cdk5 ablation would be more useful for testing these possibilities.

The negative effects of p35/p25 depletion were unexpected in light of the attenuating effects of cdk inhibitors in npc^−/− mice (Zhang et al., 2004). Because roscovitine and olomoucine are not specific for cdk5, one explanation may be that other roscovitine- and olomoucine-sensitive cdks are involved in NPC. In independent studies, we found that cell division cycle 2 (cdc2) and cdk4 are also activated in human NPC (Bu et al., 2002b) and npc^−/− mice (B. Bu and I. Vincent, unpublished data). These cdks are also present at early stages of AD (Vincent et al., 1997; Busser et al., 1998) and ALS (Nguyen et al., 2003). Roscovitine and olomoucine are most potent against cdc2 and cdk5 and essentially ineffective against cdk4 (Vesely et al., 1994; Meijer et al., 1997). Together with the present study casting doubts on cdk5, cdc2 may be the likely candidate.

An alternative but less likely explanation for the attenuating effects of cdk inhibitors in npc^−/− mice is that other related kinases such as the extracellular-regulated kinases erk1/2 or GSK3β may be important. Roscovitine and olomoucine inhibit these kinases at >10-fold the IC₅₀ for cdc2 (Meijer et al., 1997). The kinases have also been implicated in AD (Iqbal et al., 2005) and NPC (Sawamura et al., 2001). Although we showed that phosphorylation of known substrates for these kinases was not reduced by cdk inhibitors in npc^−/− mice (Zhang et al., 2004), they cannot be ruled out entirely. npc^−/− and npc^−/−, p35^−/− mice are suitable models for exploring these possibilities.

Our p35 null mice have been useful for examining the role of cdk5/p25 in other types of neurodegeneration. Induction of ischemic stroke in our p35^−/− mice has shown that cdk5/p25 is essential for the rapid or excitotoxic component of ischemic death but not the delayed or apoptotic component that requires cdk4 (Rashidian et al., 2005). Moreover, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-induced calpain-mediated p25-dependent activation of cdk5, dopaminergic neuron loss, and behavioral deficits are attenuated in the p35^−/− mice, indicating that cdk5/p25 is essential for Parkinson-type neurodegeneration (Smith et al., 2006). Conversely, the role of cdk5 and p25 in AD, ALS, and NPC remains controversial. Although some have reported increased levels of p25 in these diseases (Patrick et al., 1999; Nguyen et al., 2001; Bu et al., 2002a), others have found no increases (Yoo and Lubec, 2001; Tandon et al., 2003). At least three models of p25 overexpression have failed to result in hyperphosphorylation or relevant pathology (Ahlijanian et al., 2000; Takashima et al., 2001; Bian et al., 2002). Potential problems with overexpression models lie in the difficulties in matching activity levels to those observed in human diseases and in controlling the specific neuronal or intraneuronal distribution of increased activity. Moreover, increased cdk5 or p35 expression does not accurately simulate AD, ALS, or NPC. In contrast, the npc^−/− mouse model of neurodegeneration recapitulates the complexity of normal and pathological biochemical interactions in spatial and temporal aspects relevant to the disease. Our approach of targeted inhibition of specific effectors in such a model is a powerful means for molecular dissection of disease. Significantly, a similar approach using p35 deletion in the superoxide dismutase 1 transgenic mouse model of ALS also showed no effect on onset or progression of ALS (Takahashi and Kulkarni, 2004). These negative effects of p35/p25 depletion on NPC and ALS make it unlikely that p35/p25 would be essential for similar pathology in AD. The proposition that therapeutic management of these diseases might be achieved through targeted inhibition of cdk5 (Lau et al., 2002; Cruz and Tsai, 2004) needs to be reevaluated. The npc^−/− mouse offers the opportunity for investigating other kinases or players in these tau-associated pathologies.