Address correspondence and reprint requests to A. Tandon, Centre for Research in Neurodegenerative Diseases, University of Toronto, 6 Queen's Park Crescent West, Toronto, Ontario, Canada M5S 3H2. E-mail: firstname.lastname@example.org
Elevated levels of p25 and constitutive activation of CDK5 have been observed in AD brains. This has led to the hypothesis that increased p25 levels could promote neurofibrillary tangles (NFT) through CDK5-mediated hyperphosphorylation of tau, the principal component of NFTs. We examined p25 immunoreactivity in brains from sporadic and familial AD cases, as well as other neurologic diseases that exhibit NFT, such as Down's syndrome (DS), Pick's disease (Pick), corticobasal degeneration (CBD), progressive supranuclear palsy (PSP), frontotemporal dementia (FTD). Neither the p25 immunoreactivity nor the p25/p35 ratio was elevated in the AD brains or in the other tauopathies (n = 34) compared with controls (n = 11). Although Aβ peptides have been suggested to activate calpain-mediated cleavage of p35 to p25 in cultured neurons, p25 levels in brains of TgCRND8 mice, which express high levels of brain Aβ peptides, were similar to those of non-Tg littermates. Our data suggest that high Aβ levels in brain do not activate p35 proteolysis, and p25 is unlikely to be a causative agent for NFT formation in AD or other tauopathies.
To explore the potential role of p25 in tau-related dementias, we have examined CDK5 and p25 immunoreactivity in 45 human brains and in a mouse model of AD, TgCRND8. We report that brain p25 levels were not elevated in AD or other neurodegenerative disorders, or in TgCRND8 transgenic mice.
Materials and methods
Frozen sections of brains from Caucasians without dementia (n = 11) or with neurodegenerative disorders that were diagnosed on the basis of clinical history and postmortem neuropathological hallmarks (n = 34) were obtained from the Canadian Brain Tissue Bank (Table 1). Each brain was qualitatively categorized into one of four groups (none, mild, moderate, severe) depending on the extent of neurofibrillary tangle pathology assessed by postmortem morphological examination. Distributions of group size, sex, and mean age ± SEM were: control (n = 11; eight male, three female; 71.7 ± 1.4 year), all AD cases (n = 22; 11 male, 11 female; 68.8 ± 3.8 year), other tauopathies (n = 10; seven male, three female; two Down's syndrome, two Pick's disease, two corticobasal degeneration, two progressive supranuclear palsy, two frontotemporal dementia; 63.0 ± 3.2 year), Parkinson's disease (n = 1; 77 year), and Huntington's disease (n = 1; 61 year). Apolipoprotein E genotyping was done as described previously (Saunders et al. 1993), using DNA extracted from brain tissue. Details on the production and characterization of TgCRND8 mice (C3H X C57 genetic background) have been published elsewhere (Janus et al. 2000; Chishti et al. 2001).
Table 1. Details of frozen sections of brains from Caucasians without dementia (n = 11) or with neurodegenerative disorders that were diagnosed on the basis of clinical history and postmortem neuropathological hallmarks (n = 34) obtained from the Canadian Brain Tissue Bank
Frozen brain sections from humans or from mice that had been perfused with phosphate-buffered saline (with 1 mm EGTA) were homogenized in five volumes (g/mL) of a lysis buffer composed of: 50 mm Tris pH 7.4, 150 mm NaCl, 1% Triton-X100, 10% glycerol, 5 mm EDTA, 1 mm EGTA, 1 mm dithiothreitol (DTT), and a broad spectrum protease inhibitor cocktail (Sigma) (Patrick et al. 1999). The lysates were cleared by centrifugation at 15 000 × g for 20 min and the resulting pellet was used to prepare sarkosyl-insoluble fractions as described previously (Greenberg and Davies 1990). The soluble fractions (supernatants) were equalized to 2 mg/mL with tricine sodium dodecyl sulfate (SDS) sample buffer (Novex), as determined by BCA Protein Assay (Pierce). Samples were boiled for 3 min and equal amounts of protein (40 µg) loaded onto precast 10–20% tricine-polyacrylamide gradient gels (Novex).
Separated proteins were transferred onto nitrocellulose (0.1 µm pore size Protran, Schleicher and Schuell) for western blotting (ECL Plus, Amersham) using the following antiserum: actin (monoclonal antibody AC-40, 1 : 1000, Sigma), glyceraldehyde-3-phosphate dehydrogenase (monoclonal antibody 6C5, 1 : 1000, Biodesign International), phosphorylated tau (monoclonal antibody PHF-1), human APP (monoclonal 6E10, 1 : 2000, Senetek Inc.), CDK5 (monoclonal antibody DC-17, 1 : 1000, Santa Cruz), p35 (polyclonal antibody C-8, 1 : 1000, Santa Cruz), p25 (polyclonal antiserum C-19, 1 : 1000, Santa Cruz), polyclonal anti-p35 C-terminal antiserum (1 : 1000) kindly provided by Dr Li-Huei Tsai (Harvard), and polyclonal anti-p35 C-terminal and p25 N-terminal antiserum (1 : 1000) kindly provided by Dr Koichi Ishiguro (Mitsubishi Kasei Institute of Life Sciences). Band intensities were quantified with a Storm 860 fluorescent imager (Molecular Dynamics), which assigned arbitrary values to 100-µm square pixels relative to background fluorescence to using ImageQuant software from Molecular Dynamics. Statistical comparisons between control and disease brains were done with GraphPad InStat software using two-tailed non-parametric test (Mann–Whitney U-test) for comparisons between two groups or a non-parametric anova (Kruskal–Wallis test) for multiple comparisons.
CDK5, p35, and p25 in tauopathies and in AD
To investigate a possible role for CDK5/p25 in the development of NFT, we examined CDK5, p35, and p25 levels in homogenates of frontal cortices from eight non-neurologic controls and 14 tauopathy cases (four AD, two DS, two Pick, two CBD, two PSP and two FTD (Table 1). Considerable variation was observed within both groups, but there were no statistical differences in the levels of soluble p25 (p > 0.36), p35 (p > 0.26), CDK5 (p > 0.81) when immunoreactivities were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Fig. 1a,b). As in previous reports (Patrick et al. 1999; Takashima et al. 2001), p35 migrated as a doublet in some samples, possibly due to differential phosphorylation (Saito et al. 2003). However, the p25/p35 ratio was not significantly altered between control and tauopathy brains [2.5 ± 0.7 (range: 0.1–6.2) and 2.3 ± 0.5 (range: 0.3–5.1), respectively]. Furthermore, in contrast to a previous report (Patrick et al. 1999; Tseng et al. 2002), it was noticeable that six out of eight control brains had substantial levels of p25.
To assess whether there is evidence for altered CDK5, p35, and p25 levels in specific subtypes of AD, we measured expression of these proteins in 11 control, 13 familial (FAD) and 11 sporadic AD (SpAD) brains (Fig. 2a, soluble). Relative to the control brains, soluble CDK5 immunoreactivity in the FAD brains was similar (95%; p > 0.2), but was reduced (49%; p < 0.001) in the SpAD brains (Fig. 2b). Three anti-p35 polyclonal antisera, which were comparable in their detection of p25 and p35, showed that levels of p25 (p > 0.64 control versus FAD; p > 0.06 control versus SpAD) and p35 (p > 0.42 control versus FAD; p > 0.69 control versus SpAD) were not significantly different between control, FAD, and SpAD brains when normalized with GAPDH immunoreactivity (Fig. 2b). The ratio of p25 to p35, an indicator of p35 to p25 conversion and of the potential interaction of CDK5 with available p35 or p25, was similar between the control and both AD populations (Fig. 2c). Although, we were unable to determine with confidence sarkosyl-insoluble p35/p25 immunoreactivity, we confirmed the clear presence of sarkosyl-insoluble phosphorylated tau detected by PHF1 antibody in these FAD and SpAD cases but not in the control brains (Fig. 2a, insoluble). Significant levels of CDK5 were recovered in the sarkosyl-insoluble fraction in all control and neurological samples; although levels were quite variable, there were no differences between control and AD groups. Moreover, there was no obvious relationship between levels of insoluble CDK5 and PHF tau.
Tsai and colleagues have proposed that p35 proteolysis can occur as a postmortem artifact, and that p25/p35 ratios may be sensitive to postmortem delay (Patrick et al. 2001; Tseng et al. 2002). To investigate this, we plotted the p25/p35 ratio for the brains described in Table 1 versus the corresponding postmortem delay. No correlation was apparent (p = 0.75; Fig. 2d). Furthermore, the postmortem interval between groups was not different (p = 0.25; Kruskal–Wallis non-parametric anova).
We also considered a possible link between p25 levels and apolipoprotein E (ApoE) genotype, because carriers of the ApoE ε4 allele are at greater risk for AD (Saunders et al. 1993) and increased isoform-specific binding of ApoE ε4 with tau potentially implicates ApoE ε4 in NFT formation (Huang et al. 1994; Strittmatter et al. 1994; Huang et al. 1995). There was no relationship between p25 levels and the APOE ε4 allele (Table 1). The p25/p35 ratio in the 14 patients with at least one ε4 allele (1.7 ± 0.2) was not significantly different from that in individuals with no ε4 allele (2.4 ± 0.3).
To explore the possibility of anatomical variations in p25 levels, we examined CDK5, p35, and p25 immunoreactivity in the frontal and temporal cortex, and the cerebellum of individuals with (n = 3) and without AD (n = 3) (Fig. 3a). Some immunoreactive bands with intermediate mobilities can also be seen in the AD brain samples and may represent some transitional proteolytic derivatives of p35. Additional bands migrating between p35 and p25 were also observed in previous studies (Patrick et al. 1999; Taniguchi et al. 2001). For each brain, the intensity of each protein was normalized to the value in the frontal cortex to allow comparison between brain regions. Although there was considerable variation between individuals, intrasubject levels of CDK5, p25, p35 levels did not vary in relation to brain region (Fig. 3b).
CDK5, p35, and p25 in transgenic CRND8 mice
Acute exposure to Aβ peptides has been reported to induce calpain-mediated cleavage of p35 to p25 in cultured primary mouse neurons (Lee et al. 2000). To test whether this mechanism is recapitulated in vivo with chronically high Aβ levels as in AD (Lue et al. 1999; McLean et al. 1999; Wang et al. 1999), we examined p35 processing in a TgCRND8 transgenic mouse model of AD. Total extractable brain Aβ in these mice, which express a double mutant form of human βAPP (KM670/671NL + V717F), increases dramatically from birth and Aβ42 levels surpass Aβ40 by 8 weeks (Janus et al. 2000; Chishti et al. 2001). Figure 4(a) shows the unchanging expression of full-length human APP (APP-FL) in hemizygous TgCRND8 mice from 5 weeks to 8 months of age and an age-related rise in Aβ peptides, as compared with their non-Tg littermates. p35, p25, and CDK5 immunoreactivity did not vary in either the Tg or non-Tg animals, or with increasing age and Aβ levels (Fig. 4b). Identical results were obtained using two other anti-p35 polyclonal antisera and with six additional TgCRND8 animals aged 2–12 months (results not shown).
In the present study, we examined whether p35 processing is altered in AD or other neurodegenerative diseases exhibiting pathological accumulations of phosphorylated tau in CNS neurons, including DS, Pick, CBD, PSP, and FTD. There was no correlation between p25 levels and the presence of NFT using three different polyclonal anti-p35 antisera. Furthermore, in contrast to previous reports (Patrick et al. 1999; Tseng et al. 2002), there was no elevation of p25 levels even in the subset of brains with pure AD. In fact, p25 immunoreactivity tended to be lower in the AD brains when normalized to GAPDH levels, although the change did not reach statistical significance. Moreover, we did not detect substantial regional differences in individual brains. Interestingly, we found that detergent extractable CDK5 levels were reduced in sporadic AD brains relative to control and FAD brains, whereas insoluble CDK5 immunoreactivity was preserved. The functional implications of this change are unclear, because neither the soluble nor insoluble CDK5 levels correlated with phosphorylated tau. However, the absence of change in the insoluble CDK5 levels in the AD brains is consistent with the lack of change in p25 levels or the p25/p35 ratio, because interaction of p25 with CDK5 would be expected to reduce the amount of CDK5 complex associated with the particulate fraction (Patrick et al. 1999; Kusakawa et al. 2000).
Our results differ from the findings of Tsai and colleagues (Patrick et al. 1999) who initially reported barely detectable p25 immunoreactivity in control brains, but a 20–40-fold increase in seven out of eight AD brains. Their subsequent analysis with larger sample size yielded mean p25/p35 ratios that were approximately 1.7 and 5 for control and AD brains, respectively (Tseng et al. 2002); however, there was considerable overlap of p25/p35 ratios for the majority of the AD and control brains, ranging from 1.3 to 8. In our samples, the average p25/p35 ratio was 2.4 (range = 0.8–3.8) and 2.1 (range = 0.3–5.7) for the control and AD cases, respectively. The reasons for the disparity in the AD p25/p35 ratios between studies are not immediately apparent. It is unlikely that the p35 to p25 conversion in our samples occurred during tissue preparation, because we have used the same protocol used by Tsai and colleagues (Patrick et al. 1999). The argument that prolonged postmortem delays expose p35 to proteases that are activated in dying neurons (Patrick et al. 2001; Taniguchi et al. 2001) does not account for the different results because the mean postmortem interval for the control brains used in our study (5.7 ± 1.0 h) was substantially less than the 21 ± 1.7 h reported by Patrick et al. (1999), but similar to that reported by Tseng et al. (2002) (∼3 h). Indeed, the p25/p35 ratios in the present study are comparable to the control values obtained in the Tseng et al. (2002) analysis. Moreover, we observed no correlation between the p25/p35 ratio and the corresponding postmortem delay. Our results are more compatible with other reports showing control human brains expressing p25 in excess of p35, and no difference in the p25/p35 ratio relative to AD brains (Takashima et al. 2001; Taniguchi et al. 2001; Van den Haute et al. 2001; Yoo and Lubec 2001).
The neurotoxicity following acute Aβ peptide treatment of cultured primary mouse neurons has been proposed to involve calpain-mediated proteolysis of p35 to p25 (Lee et al. 2000; Li et al. 2003). The p25 derivative is more stable than its parent peptide when expressed artificially, and can sequester CDK5 in the cell soma and up-regulate its kinase activity (Patrick et al. 1998; Patrick et al. 1999; Ahlijanian et al. 2000). We have explored whether this cascade can be extended to an in vivo setting using TgCRND8 mice, which express a double mutant form of βAPP. TgCRND8 mice accumulate Aβ in the brain, which is accompanied by Aβ plaques and severe cognitive deficits at three months of age (Janus et al. 2000; Chishti et al. 2001). Despite the significant age-related rise in Aβ peptide levels (especially of Aβ1−42) between 1 and 8 months, there was no concomitant change in p35 or p25 levels. This suggests considerable divergence between the cell culture and in vivo models of Aβ toxicity with respect to p35 metabolism. While this result could be dismissed because of the absence of NFT in this and other Tg mice overexpressing Aβ, it is worth noting that the acute toxicity studies showing Aβ-induced calpain activation and p35 cleavage were also performed on murine neurons. There also appear to be important differences in p35 metabolism in mouse and human neurons; comparison of p25/p35 ratios, presented in this study and in previous reports (Lee et al. 2000; Taniguchi et al. 2001; Van den Haute et al. 2001; Bian et al. 2002; Tseng et al. 2002), suggests that murine cells have more p35 than p25 (p25/p35 ratio < 1), whereas human cells contain more p25 than p35 (p25/p35 ratio > 2). Whether this is strictly a species-related phenomenon or whether it reflects some early postmortem changes is currently unclear, and needs to be resolved. On balance, the TgCRND8 mouse model is a more representative model of Aβ biology in AD patients than acute studies with cultured neurons. Furthermore, the absence of increased p25 levels in the transgenic mouse model is consistent with the lack of p25 accumulation in our collection of AD brains. When taken together, our results suggest that cleavage of p35 to p25 is unlikely to contribute significantly to AD pathology.
This work was supported by grants from the Canadian Institutes of Health Research Howard Hughes Medical Institute, Alzheimer Society of Ontario, Canadian Genetic Diseases Network, and the Scottish Rite Foundation.