Physiological and pathological phosphorylation of tau by Cdk5

1Alzheimer’s disease (AD) is the most common neurodegenerative dementia and affects more than 35 million people worldwide. Thus, the development of therapeutic methods is urgently needed to determine the underlying molecular mechanism of AD. Major pathological hallmarks of AD include senile plaques and neurofibrillary tangles (NFT), which consist mainly of amyloid β peptide (Aβ) and hyperphosphorylated tau, respectively. Mutations of the amyloid precursor protein (APP) and presenilin, a component of γ-secretase, are found in familial AD, and previous studies have established the hypothesis of the amyloid cascade. On the basis of this hypothesis, great effort has been paid to develop drugs to reduce Aβ production or to clear Aβ, but successful results have not yet been obtained. In contrast, it has been shown that tau pathology is more closely related to neuronal loss. Tau is a genetic factor of a neurodegenerative disease known as frontotemporal dementia parkinsonism linked with chromosome 17 (FTDP-17;;; ). FTDP-17 tau mutants are highly phosphorylated in patient brains. Regardless of whether phosphorylation is a cause of FTDP-17, it is still critical to determine the neuronal milieu in which tau hyperphosphorylation occurs. Cyclin-dependent kinase 5 (Cdk5) is a major tau kinase that is involved in abnormal phosphorylation in AD brains. Here, we summarize the phosphorylation of tau by Cdk5. To the best of our knowledge, this is the first review article focused specifically on Cdk5 phosphorylation of tau.

2Tau is a member of the heat-stable microtubule-associated proteins (MAPs), which consist of MAP2 and MAP4. Tau, as well as MAP2, is mainly expressed in mammalian neurons. While MAP2 is localized in dendrites, tau binds to microtubules that are present in axons and is thus often used as an axonal marker. Similar to MAP2, tau binds to microtubules via the C-terminal microtubule-binding repeats, which consist of three or four imperfect repeats of 31 or 32 amino acids. There are six isoforms in human tau that are dependent on the presence or absence of one or two N-terminal insertions and a C-terminal region with three (3R) or four (4R) microtubule-binding repeats. The longest isoform of human tau is composed of 441 amino acids; the phosphorylation sites and mutation sites are usually numbered according to this isoform of tau (Figure 1). We also used this notation in this article. Physiologically, tau promotes microtubule assembly and stabilizes microtubules by laterally binding to the surface of microtubules. In addition to the classical functions, new functions in signaling and cytoskeletal organization have emerged. These activities are regulated by phosphorylation in the microtubule-binding repeat or the franking region by a number of protein kinases. Thus, the physiological function of tau is regulated by phosphorylation. Tau is a naturally unfolded protein with an extended structure but aggregates into NFTs in the brains of AD patients. A number of neurodegenerative diseases with tau aggregates are collectively known as tauopathy. Tau in aggregates is hyperphosphorylated, and this hyperphosphorylation is a feature employed for the diagnosis of diseases. However, it is not completely known how pathological tau is hyperphosphorylated and what role hyperphosphorylation plays in aggregate formation and disease development.

3Cdk5 was first purified as one (TPKII) of two tau protein kinases, namely TPKI and TPKII, from a bovine brain microtubule fraction, as neuronal cdc2-like kinase (nclk) from bovine brain extracts, and as a KSP sequence phosphorylating kinase from rat spinal cord. Purified Cdk5 is a complex that consists of the Cdk5 catalytic subunit with a molecular mass of approximately 31 kDa and a polypeptide with a molecular mass of approximately 23–25 kDa (p25). p25 was subsequently shown to be an N-terminal truncated form of its full-length version of p35. Cdk5 has attracted attention as a potential disease tau kinase because Cdk5 phosphorylated tau at sites that were hyperphosphorylated in AD brains. Importantly, there is an accumulation of p25 in AD brains with higher phosphorylation ability of Cdk5-p25 for tau compared Cdk5 activated by p35, which highlights the importance of Cdk5-p25 in abnormal tau phosphorylation.

4Cdk5 is a member of the Ser/Thr cyclin-dependent kinases (Cdks). Cdk5 is a catalytic subunit and is activated by binding to its regulatory subunit, p35 or p39 (Figure 2). Cdk5 has a 55~60% amino acid sequence homology to well-known cell cycle Cdks, such as Cdk1, 2, 4, and 6, but its activators, namely p35 and p39, display no homology to cyclins, which are activators of cell cycle Cdks. p35 and p39 consist of 307 and 369 amino acids, respectively, with a Cdk5 activation domain in the C-terminal region. The crystal structure has revealed that the activation domain of p35 has a tertiary structure, which resembles cyclin A in the Cdk2-cyclin A complex, that explains its ability to activate Cdk5. Although cell cycle Cdks are activated at a particular cell cycle phase in proliferating cells and promote cell cycle progression, Cdk5 is mainly active in post-mitotic neurons, although Cdk5 is expressed widely in many cells types and tissues. This is because p35 and p39 are predominantly expressed in neurons. The number of reports describing the kinase activity of Cdk5 in extra-neuronal cells or tissues is increasing, but the activation mechanism and function are poorly understood yet. While p35 could be the major activator, cyclin I was indicated to be an activator of Cdk5 in kidney podocytes.

5Cdks are a family of proline-directed protein kinases (PDPKs) that phosphorylate Ser or Thr residues followed by proline, that is, Ser/Thr-Pro sequences. The substrate specificity of Cdk5 is very similar to that of Cdk1-cyclin B. There are at least three different PDPKs, namely MAP kinases such as ERK1/2, GSK3, and Cdks, but their substrate preferences are slightly different. Cdk5 prefers the (Ser/Thr)-Pro motif with a basic amino acid at the second C-terminal site, (Ser/Thr)-Pro-X-(Lys/Arg) (X is any amino acid;,;; ). ERK1/2 phosphorylates Pro-X-(Ser/Thr)-Pro sequences, and GSK3β can phosphorylate (Ser/Thr)-(X)3-(pSer/pThr) with priming phosphorylation at +4 site in addition to (Ser/Thr)Pro sequences.

6In contrast to cell cycle Cdks, Cdk5 does not require phosphorylation of the Cdk5 subunit in the activation loop for activation. Binding of the regulatory subunits p35 and p39 is sufficient for activation. The kinase activity of Cdk5 is mainly determined by the available protein amounts of p35 or p39 in neurons. p35 and p39 are proteins with a short life in which the half-life is about 30 min and 120 min for p35 and p39, respectively. p35 and p39 are degraded by the ubiquitin-proteasome system as are other short half-life proteins. Thus, one major factor that regulates Cdk5 activity is degradation of p35 and p39 via proteasomal degradation. Phosphorylation at Thr138 in p35 by Cdk5 stimulates this degradation. Because phosphorylation decreases with aging, the half-life of p35 becomes longer in adult mouse brains; however, it is not known how the phosphorylation state is determined or the identity of the E3 ubiquitin ligase for p35.

7The cellular localization is also different between cell cycle Cdks and Cdk5. While cell cycle Cdks are mainly present in the nucleus, where they promote the cell cycle, active Cdk5 associates with membranes in the cytoplasm. This is achieved by the membrane binding properties of p35 and p39, which are myristoylated at the N-terminal Gly (Figure 2;;;, ). There are excess amounts of Cdk5 compared to p35 and p39, and these free Cdk5 molecules are soluble. The binding of p35 or p39 not only activates Cdk5, but it also induces its membrane association. Membrane association is also supported by Lys residues in the N-terminal p10 region of p35 and p39 (+charge in p35 in Figure 2). This membrane association may not only restrict its targets to proteins localized to the submembranous regions but also prohibits its nuclear translocation.

8When neurons suffer from stress, death signals or overexcitation, large influxes of Ca2+ enter into the cytoplasm, resulting in the activation of the calcium-dependent protease calpain. Calpain cleaves p35 to p25, an N-terminal truncated form that consists of the C-terminal Cdk5 activation domain (Figure 2;;; ). Subsequently, in contrast to Cdk5-p35, Cdk5-p25 is released from the membranes and is capable of accessing proteins. In addition, a component of Cdk5-p25 is known to enter into the nucleus to activate the cell cycle machinery. Furthermore, Cdk5-p25 acquires a longer half-life, resulting in the net activation of Cdk5.

9Tau is phosphorylated at approximately 45 sites in AD brains (Figure 1). Phosphorylation characteristically occurs at many (Ser/Thr)-Pro sequences (Figure 1;; ). Among the 16 (Ser/Thr)-Pro sequences in tau, Cdk5 phosphorylates 9–13 sites. However, the reported sites are not always the same. Initially, Ishiguro’s group identified Ser202, Thr205, Ser235, and Ser404 as TPKII (Cdk5-p25) phosphorylation sites using amino acid sequencing. In addition, using purified nclk (Cdk5-p25), reported Ser195, Ser202, Thr205, Thr231, Ser235, Ser396, and Ser404. demonstrated Ser202, Thr205, Ser235, and Ser404 to be major sites with Thr153 and Thr212 as minor sites in in vitro Cdk5-p25 phosphorylated tau using 2D-phospho-peptide mapping and mass spectrometry analysis. Phosphorylation sites of human tau via in vitro recombinant Cdk5-p20 (shorter activation construct than p25) include Thr181, Thr205, Thr212, Thr217, Ser396, and Ser404, which were determined using mass spectrometry. reported that Cdk5-p25 phosphorylates Tau at Thr181, Ser199, Ser202, Thr205, Thr212, Ser214, Ser217, Thr231, Ser235, Ser396, and Ser404 using a repertoire of phospho-specific antibodies. Using NMR,, ) analyze tau phosphorylation by Cdk2-cyclin A3 and Cdk5-p25; while Cdk2-cyclin A3 phosphorylated Thr153, Ser199, Ser202, Thr205, Thr231, Ser235, and Ser404, with high levels of phosphate incorporation at Ser202/Thr205 and Thr231/Ser235, Cdk5-p25 needed GSK3β for the same phosphorylation profile with Cdk5-p25 providing Ser202, Thr205, Ser235, and Ser404 as major sites. We believe that these differences in phosphorylation sites were due to the methods and kinase preparations used for analysis. The use of phospho-specific antibodies may detect minor phosphorylation sites, which are sometimes below the detection level of biochemical methods. Further, immunoblotting with phospho-specific antibodies has a problem in quantification of multiply phosphorylated proteins. Purification or in vitro reconstruction of active Cdk5 is also challenging. If the kinase activity is not sufficiently high, then the contribution of contaminating kinases may present challenges.

10We have determined the phosphorylation sites of tau using 2D phosphor-peptide mapping in vitro (Cdk5-p25 purified from porcine brains) and in cultured cells (co-transfected Cdk5-p35 or Cdk5-p25) and primary neurons (endogenous Cdk5-p35) using isotope labeling methods. The major in vitro Cdk5 phosphorylation sites determined were Ser202 or Thr205 (spot 2), Ser235 (spot 4), and Ser404 (spots 3 and 5; Figure 3A). A similar in vitro 2D-phosphopeptide pattern has been reported by. These sites were detected as major sites in cultured COS-7 cells when Cdk5-p25 was co-transfected and in primary neurons (Figure 3A; ). These results indicated that Cdk5 is a major kinase that phosphorylates tau in cultured neurons. The signal of spot 1, which is a doubly phosphorylation spot of Ser202 and Thr205, is strong in cultured neurons but weak when phosphorylated by Cdk5 in vitro. Interestingly, Ser202 and Thr205 are exclusive phosphorylation sites for Cdk5, but only one of these sites is phosphorylated, with a preference for Ser202 by Cdk5 (Figure 3B; ). However, both sites can be phosphorylated by Cdk5 on the tau molecule when bound on microtubules. The strong signal on spot 1 in cultured neurons suggests that most tau in neurons binds to microtubules or that the sites are phosphorylated by multiple kinases.

11According to the phosphorylation sites described above, the (Ser/Thr)-Pro sequences in tau can be grouped into three categories: (1) major Cdk5 phosphorylation sites, (2) minor Cdk5 phosphorylation sites and (3) non-Cdk5 phosphorylation sites (Figure 4). As tau is a filamentous protein, the amino acid sequence around the phosphorylation sites may affect their phosphorylation ability more than those in globular proteins. Thus, several interesting aspects can be observed; three of the four major sites, specifically Ser202, Thr205 and Ser235, have Pro at the second N-terminal site, that is, Pro-X-(Ser/Thr)-Pro, which is also a consensus sequence for MAP kinase. Thus, these sites may be commonly phosphorylated by MAP kinase. Ser404 is a nearly perfect consensus of Cdk5. Four of the six minor phosphorylation sites have Lys or Arg at the N-terminal next to the phosphorylation sites: (Lys/Arg)-(Ser/Thr)-Pro. These Lys-Ser-Pro (KSP) sequences are phosphorylation sites found in neurofilament H and M subunits, and phospho-specific antibodies for these sites are known to react with phospho-tau. In contrast to these phosphorylation sites, there are many acidic amino acids around non-Cdk5 phosphorylation sites (labeled in blue in Figure 4), except for Thr175. These acidic amino acids may decrease the propensity of phosphorylation. Thus, Cdk5 cannot phosphorylate the (Ser/Thr)-Pro site if it is already phosphorylated. This finding is consistent with the exclusive relationship between Ser202 and Thr205 phosphorylation and may also explain why Ser231 and Ser199 are minor phosphorylation sites, as they have a better Cdk5 phosphorylation site close to them. This is in contrast to GSK3β, which requires priming phosphorylation around the phosphorylation sites (Figure 1; ).

12The Cdk5 phosphorylation sites described above are the results of in vitro or cultured cell experiments. However, we believe that these properties can be applied to in vivo phosphorylation. As described above, Ser404 appears to be a Cdk5-specific site. Ser404 is one of the most phosphorylated sites in rat brain independent of whether the phosphorylation was examined using mass spectroscopy or an anti-phospho-antibody, which suggest that Cdk5 is a major tau kinase in normal brains. Indeed, knockdown of Cdk5 reduced tau phosphorylation at the PHF-13 site (Ser396/Ser404; ), and transgenic expression of p25 in mouse brain increases tau phosphorylation at several sites, including AT8 and PHF-1 sites. In contrast, several reports argue against in vivo Cdk5 phosphorylation of tau. For example, phosphorylation of tau is not decreased in mice with a null mutation of p35. Furthermore, an increase in overall tau phosphorylation was not detected in p25 transgenic mice, although phosphorylation at specific sites such as Ser202 and Ser235 was shown to increase in another p25 mouse model. Tau expressed in yeast lacking the Cdk5 homolog Pho85 showed an increase, not a decrease, in phosphorylation. These reports demonstrate GSK3β to be the predominant tau kinase and its inhibition via Cdk5. However, one challenge is the lack of methods used to specifically detect Cdk5 phosphorylated sites. Phospho-specific antibodies have been used. The most frequently used antibodies are AT-8 and PHF-1; however, these antibodies require multiple phosphorylation for immunoreaction; for example, for AT-8, it requires phosphorylation at Ser199 in addition to Ser202 or Thr205 phosphorylation by Cdk5; for AT180, a recent report indicates AT180 requires only Thr231 phosphorylation although it was often used as evidence for phosphorylation at both Thr231 and Ser235; and for PHF-1, it requires phosphorylation at Ser369 following Ser404 phosphorylation by Cdk5. Thus, specific tools for each Cdk5 site alone are required.

13It is generally considered that phosphorylation of tau by Cdk5-p35 is physiological and that phosphorylation by Cdk5-p25 is pathological. Most in vitro experiments have been done using Cdk5-p25 because purified Cdk5 is complexed with p25. Thus, these results should include information on abnormal phosphorylation of tau. Nevertheless, the entity of hyperphosphorylation is still unclear; is there an increase in the number of phosphorylation sites, an increase in the extent of phosphorylation at these particular sites, or both? Two biochemical studies have addressed the kinetics of tau phosphorylation using recombinant Cdk5-p35 and Cdk5-p25. However, both studies arrived at different conclusions. One study showed a higher affinity of Cdk5-p25 for tau compared to Cdk5-p35, while the other did not obtain similar findings. Considering the multiple phosphorylation sites in tau with different amino acid sequences surrounding the sites, it may be difficult to determine the kinetic parameters of tau phosphorylation in vitro. Cellular phosphorylation may be even more complicated and is affected by the accessibility of Cdk5 activated by p35 and p25, phosphorylation by other kinases and the binding of tau to microtubules. Our simple comparison using 2D-phospho-peptide mapping indicates that the major phosphorylation spots were similar between tau phosphorylated by Cdk5-p35 and Cdk5-p25. On the basis of these results, we propose that Cdk5-dependent abnormal phosphorylation represents an increase in the rate of phosphorylation at the particular sites rather than an increase in the number of sites.

14An increase in phosphorylation by Cdk5 would further elevate the total phosphorylation of tau by facilitating subsequent phosphorylation with GSK3β. GSK3β is another tau kinase, which is also known as TPKI. GSK3β phosphorylates Ser/Thr residues with priming phosphorylation at the +4 site, (Ser/Thr)-X3-(pSer/pThr). In fact, GSK3β phosphorylation is considerably enhanced by prior phosphorylation with Cdk5. Although many GSK3β sites have been reported, the major sites are Ser199, Ser202, Thr231, Ser396, Ser400, and Ser412. Thr 231 and Ser396/Se400 are primed by phosphorylation at Ser235 and Ser404 by Cdk5, respectively. Thus, Cdk5 would increase the total phosphorylation of tau with GSK3β in an additive manner.

15The net phosphorylation is a result of the balance between phosphorylation and dephosphorylation. Hyperphosphorylation should be attained by either increased phosphorylation or decreased dephosphorylation. What is the contribution of dephosphorylation? As previously reported, protein phosphatase 2A is a major phosphatase for tau. The dephosphorylation velocity of tau at Cdk5 sites is slower compared to PKA phosphorylation sites, which may, at least in part, be due to the stronger resistance of phosphorylation at the (Ser/Thr)Pro sites against PP2A-dependent dephosphorylation compared to other sites. Dephosphorylation at the (Ser/Thr)Pro sites is modulated by Pin1 peptidyl-prolyl cis/trans isomerase. Pin1 changes the conformation of the peptide bond at proline from cis to trans, and the trans-conformation is easily dephosphorylated by PP2A (Figure 5). Dephosphorylation of Cdk5 phosphorylation sites at Ser202 and Ser235 is delayed in the absence of Pin1. The contribution of Pin1 in tau aggregate formation in AD has been demonstrated in Pin1-deficient mouse brains. In addition to the four major Cdk5 sites, dephosphorylation at Ser212 and Thr231 is also stimulated by Pin1, although it is recently shown that dephosphorylation of Thr231 is not Pin1-dependent. There may be more (Ser/Thr)-Pro phosphorylation sites in which dephosphorylation is regulated by Pin1. We also demonstrated that phosphorylation of FTDP-17 mutant tau, P301L and R406W, by Cdk5 exhibits a slightly weaker affinity to Pin1 compared to WT tau, which may result in decreased dephosphorylation of mutant tau by PP2A, consistent with previous reports that the FTDP mutation demonstrates weakened binding to PP2A.

16Frontotemporal dementia with parkinsonism linked to chromosome 17 is a dominantly inherited disease of neurodegenerative dementia with mutations in the tau gene MAPT. There are more than 40 FTDP-17 mutations reported, most of which are missense mutations found in the microtubule binding repeats with four mutations at the N-terminal and six in the C-terminal regions (Figure 1; ). FTDP-17 mutant tau forms aggregates in the frontal cortex of patient brains and tau is abnormally phosphorylated at the sites of phosphorylation in AD. Thus, it is thought that these mutations are highly involved in abnormal phosphorylation. Phosphorylation of FTDP-17 mutants has been studied predominantly in mice overexpressing mutant tau. Overexpressed mutant tau forms sarkosyl-insoluble aggregates in mouse brains, and aggregates can be labeled using phospho-specific antibodies. However, similar to AD tau, it is not known how FTDP-17 mutation induces abnormal phosphorylation.

17Several groups have examined the phosphorylation of FTDP-17 tau mutants by Cdk5. showed that mutations of P301L, V337M, R406W, and G272V promoted the mobility shift of tau upon phosphorylation by Cdk5, whereas R406W inhibited Ser404 and Ser235 phosphorylation. They concluded that the increased mobility shift was due to an increase in phosphorylation at Ser202, Ser404, and Ser235 but not due to the higher extent of total phosphorylation. studied the phosphorylation of six FTDP-17 mutants, namely G272D, N279K, ΔK280, P301L, V337M, and R406W, in yeast strains lacking mds1 GSK3β homolog and/or Pho85 Cdk5 homolog. They demonstrated that P301L and R406W showed lower AD2 (phosphorylation at Ser396/Ser404) and PG5 (phosphorylation at Ser409) reactivity and that the reaction was reduced in the mds1-lacking strain and stimulated in the Pho85-deficient strain, suggesting that Ser396/Ser404 sites are phosphorylated by GSK3β and mutations affecting Ser409 phosphorylation. We have examined the phosphorylation of K257T, P301L, P301S, and R406W mutants in vitro and in cultured cells by Cdk5 using 2D-phosphopeptide mapping. The phospho-peptide patterns were identical between WT and mutant tau except for R406W. Phosphorylation at Ser404 was lacking in R406W. We propose that the R406W mutation alters the preferred consensus at the Ser404 site for Cdk5 to less preferred sequences. A distinct difference in phosphorylation was consistently reported with R406W tau, although these results were variable; reduced phosphorylation at many phosphorylation sites, reduced phosphorylation at specific sites, or greater overall phosphorylation was observed. One R406W patient showed a different progression of the disease from other FTDP-17 patients. Thus, it would be interesting to determine the role of this specific property of phosphorylation in disease development.

18Numerous reports have described an increased phosphorylation of tau when neurons suffer from various neurotoxic insults, such as Aβ, ischemia/hypoxia, oxidative stress, inflammation, and excitotoxicity. Aging may also represent a specific stress factor. Many of these studies observed an increase in p25 and argued Cdk5-dependent phosphorylation of tau. However, the increase in p25 and tau phosphorylation were performed in parallel and their direct correlation was not mostly demonstrated. Considering that Cdk5 sites can be phosphorylated by other PDPKs and that relatively minor phosphorylation sites were examined using western blotting analyses with anti-phospho-specific antibodies, further studies are required to determine their causal relationship.

19Elucidation of the molecular mechanism inducing hyperphosphorylation of tau in tauopathic brains including AD is one of the critical issues for the prevention of dementia development independent of hyperphosphorylation as a cause of disease. Hyperphosphorylation must reflect the cellular conditions of affected neurons in disease brains. Cdk5 has been extensively studied as one of the major kinases because Cdk5 generates disease-specific phosphorylation epitopes. However, despite intensive previous studies, it is still unclear how Cdk5 contributes to tau phosphorylation physiologically and pathologically. In particular, in vivo phosphorylation by Cdk5 has not been convincingly demonstrated. This may, at least in part, be due to an overlap in the phosphorylation of many (Ser/Thr)Pro sites in tau by several PDPKs. Another possibility is the lack of methods used to specifically identify Cdk5 phosphorylation. By overcoming these challenges, studies on tau phosphorylation by Cdk5 can provide valuable insight on the molecular mechanism underlying AD and the development of strategies to prevent dementia.

20The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.