Mice lacking phosphatase PP2A subunit PR61/B’δ (Ppp2r5d) develop spatially restricted tauopathy by deregulation of CDK5 and GSK3β

Abstract

Functional diversity of protein phosphatase 2A (PP2A) enzymes mainly results from their association with distinct regulatory subunits. To analyze the functions of one such holoenzyme in vivo, we generated mice lacking PR61/B’δ (B56δ), a subunit highly expressed in neural tissues. In PR61/B’δ-null mice the microtubule-associated protein tau becomes progressively phosphorylated at pathological epitopes in restricted brain areas, with marked immunoreactivity for the misfolded MC1-conformation but without neurofibrillary tangle formation. Behavioral tests indicated impaired sensorimotor but normal cognitive functions. These phenotypical characteristics were further underscored in PR61/B’δ-null mice mildly overexpressing human tau. PR61/B’δ-containing PP2A (PP2A_T61δ) poorly dephosphorylates tau in vitro, arguing against a direct dephosphorylation defect. Rather, the activity of glycogen synthase kinase-3β, a major tau kinase, was found increased, with decreased phosphorylation of Ser-9, a putative cyclin-dependent kinase 5 (CDK5) target. Accordingly, CDK5 activity is decreased, and its cellular activator p35, strikingly absent in the affected brain areas. As opposed to tau, p35 is an excellent PP2A_T61δ substrate. Our data imply a nonredundant function for PR61/B’δ in phospho-tau homeostasis via an unexpected spatially restricted mechanism preventing p35 hyperphosphorylation and its subsequent degradation.

Reversible protein phosphorylation, spatio-temporally controlled by kinases and phosphatases, is a key mechanism regulating a plethora of physiological processes. Protein phosphatase 2A (PP2A) is a major Ser/Thr phosphatase broadly implicated in cellular signaling (1, 2), and its deregulation results in pathologies such as cancer (2, 3) and Alzheimer’s disease (AD) (4). Gene knockout (KO) of the most abundant isoform of the catalytic subunit (PP2A_Cα) in mice is embryonically lethal (5), underscoring the broad physiological importance of PP2A in mammals.

The underlying basis of this functional diversity is formed by PP2A’s complex structure and regulation. PP2A enzymes are heterotrimers (PP2A_T) consisting of a core dimer (PP2A_D) encompassing a structural A and a catalytic C subunit that is associated with one regulatory B-type subunit. The B-type subunits critically control PP2A activity by regulating substrate selectivity and/or directing enzyme localization within cells (1–3). Four families of B-type subunits have been identified: PR55/B (B55), PR61/B’ (B56), PR72/B” and striatins/SG2NA/B”’. Within each family, distinct (alternatively spliced) genes encode various structurally related isoforms, which—combined with two isoforms of both A and C—results in a variety of holoenzymes, all theoretically exhibiting distinct functions (1–3). However, despite substantial efforts, specific holoenzymes involved in a particular physiological process often remain elusive, and it is far from clear whether related subunits may perform redundant functions.

In the current study, we generated KO mice deficient for the B-type subunit PR61/B’δ (also called B56δ). PR61/B’δ represents the longest isoform (74 kDa) of the PR61/B’ class (6, 7) and is ubiquitously expressed, with highest abundance in brain and embryo (7). In contrast to PP2A_Cα KO mice, mice lacking PR61/B’δ are viable without an obvious phenotype early in life. However, because of the high expression in brain in wild type (WT), a neural phenotype could be expected in the KO, and this was further analyzed. In addition to some general functional redundancies, our findings demonstrate an indirect and spatially restricted role for PP2A_T61δ in tau phosphorylation homeostasis, implying PP2A B-type subunits exert specific nonredundant functions in vivo.

Results

PR61/B’δ-Null Mice are Viable and Fertile.

Mice lacking the PR61/B’δ gene (Ppp2r5d) were generated by homologous recombination (Fig. S1). Loss of protein expression was confirmed with a specific PR61/B’δ antibody (7) in liver, brain, and heart lysates (Fig. S1). Despite high PR61/B’δ levels during embryogenesis in WT (7), PR61/B’δ^-/- mice turned out to be viable. Breeding of PR61/B’δ^+/- mice resulted in Mendelian ratios of progeny without sexual bias, and litter sizes of KOxKO and WTxWT matings were comparable (7.55 vs 6.83 pups/litter, p < 0.05). Placental sections of pregnant females from KOxKO matings showed normal decidualization and implantation with differentiated trophoblasts (Fig. S2). No significant changes were observed in expression of PR61/B’α/γ2/γ3/ϵ, PR55/Bα, PR130/B”α1, A or C subunits in total extracts of KO liver, heart, and brain at three months (Fig. S3).

Localization of PR61/B’δ in Neural Tissues and PC-12 Cells.

Immunohistochemical (IHC) stainings with the PR61/B’δ antibody (7) revealed neuronal, cytoplasmic expression in most brain areas (Fig. 1A) with highest expression in striatum (Fig. 1E) and thalamus (Fig. 1F), intermediate expression in brain stem (Fig. 1G), CA1, CA2, CA3 regions (Fig. 1H) and dentate gyrus of hippocampus (Fig. 1I), and weak expression in cortex (Fig. 1J) and cerebellum (Fig. 1K). These patterns generally match well with previous in situ hybridization data (7). No staining was detected in corresponding KO brain areas (Fig. 1B and 1L), further demonstrating antibody specificity. Expression of PR61/B’δ was also seen in spinal cord (Fig. 1C) and again absent in the KO (Fig. 1D). Immunofluorescence (IF) demonstrated PR61/B’δ localization in cytoplasm, nucleus (weak), and dendrites of neuronally differentiated PC-12 cells, along partial colocalization with F-actin, microtubuli, and tau (Fig. 1M).

Fig. 1.

PR61/B’δ localization in brain, spinal cord, and PC-12 cells. Immunohistochemistry staining (3-step procedure) of PR61/B’δ in WT brain (A) and spinal cord (C). Brain (B) and spinal cord (D) of PR61/B’δ KO mice are shown as control. Detailed expression of PR61/B’δ in WT striatum (E), thalamus (F), brain stem (G), CA1, 2 and 3 regions (H) and dentate gyrus (DG) of hippocampus (I), cortex (J), and cerebellum (K). KO cerebellum is shown as control (L). Nuclei are counterstained with haematoxylin (blue). Scale bars: 1 mm (A–D, K, and L); 50 μm (E–J). (M) Immunofluorescence showing colocalization of PR61/B’δ (green) and tau (red) in NGF-differentiated PC12 cells (Scale bar: 10 μm).

Progressive Tau (Hyper)Phosphorylation in Brain and Spinal Cord of PR61/B’δ KO Mice.

Tau binds and stabilizes microtubules depending on its phosphorylation state (8). Because PP2A is a well-established tau phosphatase in vivo (4), capable of dephosphorylating specific sites in vitro and in situ (9–12), we monitored tau phosphorylation in PR61/B’δ KO mice using phosphorylation-sensitive monoclonal antibodies AT8 (Ser202/Thr205), AT180 (Thr231), AT100 (Thr212/Ser214), and AD2 (Ser396/Ser404). These immunostainings indicated phosphorylation at AT8 and AT180 epitopes in brain stem and dorsal horn of the cervical spinal cord of six-month-old KO mice, whereas no or very little immunoreactivity was detected in age-matched WT controls (Fig. 2A and Fig. S4). In younger mice ( three months) no enhanced tau phosphorylation was observed (Fig. 2A and Fig. S4). Because it is known from transgenic models tau phosphorylation increases with age (13), we also performed IHC studies with 18-month-old mice. Aging did not only correlate with increased tau hyperphosphorylation in brainstem and spinal cord (Fig. 2A and Fig. S4), it also resulted in a broader distribution of this phenotype as these mice also displayed weak tau phosphorylation in subiculum, lateral dentate cerebellar nucleus, and cortex (very weak). Western blotting confirmed increased AT8/AT180 immunoreactivity in brain stem and spinal cord of 18-month-old KO mice, while total tau levels did not significantly change with age (Fig. 2B). Staining with AT100, recognizing a phospho-epitope preceding neurofibrillary tangle (NFT) formation (14), was positive in brain stem and spinal cord of six-month-old KO mice (Fig. 2C and Fig. S4) but did not increase with age. Similar observations were made for AD2, recognizing phospho-Ser396/Ser404 (Fig. 2C and Fig. S4). AT100 and AD2 Western blots were negative. In addition, cytoplasmic MC1 staining, recognizing a conformational tau epitope found in AD (15), was detected again in brain stem and spinal cord of six-month-old KO mice, while it decreased at 18 months (Fig. 2D and Fig. S4) and was absent in WT mice. Because tau conformation defined by MC1 highlights transition from soluble to filamentous tau (13, 15), these data indicate that depending on age, tau is in a hyperphosphorylated (AT8, AT180 and less AT100, AD2), structurally different (MC1) state in the KO. Despite these indications, tau did not aggregate into filaments or tangles in older mice because CongoRed/X34 and Bielschowsky staining did not reveal an associated NFT pathology (Fig. S4). TUNEL staining did not indicate any apoptotic cell death (Fig. S4). NFT absence might be explained by physiological clearance of MC1-positive tau by the protective chaperone-tau processing pathway (16). Chaperones HSP70 and HSP90 are indeed significantly elevated in brain stem and spinal cord of six-month-old KO mice as compared to WT, while this is not the case in older mice (Fig. S5).

Fig. 2.

Age-related tau hyperphosphorylation and misfolding in brain stem and spinal cord of PR61/B’δ KO mice. (A) Immunohistochemistry (IHC) with AT8 (Ser202/Thr205) and AT180 (Thr231) showing progressive tau phosphorylation in brain stem (left) and dorsal horn of the cervical spinal cord (right) of PR61/B’δ KO mice. (B) Western blot of brain stem and spinal cord lysates (n = 5) developed with the indicated antibodies. (C) IHC with AT100 (Ser212/Ser214) and AD2 (Ser396/Ser404) at six months in brainstem (left) and spinal cord (right). (D) IHC with MC1 (recognizing misfolded tau) in KO and age-matched WT brain stem (left) and spinal cord (right). Only cytoplasmic tau stainings were quantified (A, C, and D) with Leica QWin V2.8 software. (E) Representative IHC images from brainstem at six months (for complete overview, see Fig. S4). Scale bars: 100 μM.

Phenotypes of Tau-Overexpressing PR61/B’δ KO (TKO) Mice.

We crossed PR61/B’δ KO with hTau4 (TT4^+/-) mice (17), mildly expressing WT four-repeat human tau in all neurons (TKO mice) (Fig. S6). In contrast to mutant tau transgenic mice (13, 14), TT4^+/- mice do not show NFT pathology but a milder phenotype, characterized by progressive tau phosphorylation at specific pathological epitopes throughout the entire brain (17). Thus, we could ask whether and where deletion of PR61/B’δ in this model would be able to aggravate the TT4^+/- phenotype and potentially lead to NFT. AT8 and AT180 Western blots showed again increased tau phosphorylation specifically in hindbrain and spinal cord lysates of 10-month-old TKO mice (Fig. 3A), thus confirming that even in the presence of overexpressed tau, PR61/B’δ absence fails to result in a forebrain tau pathology. Likewise, NFT are still absent in TKO mice as witnessed by CongoRed/X34 and absence of AT8 immunoreactivity in the sarkosyl-insoluble fraction of total TKO brain lysates (Fig. 3B). The latter was only detected in Tau-P301L (TPLH) transgenic mice, as previously reported (13) (Fig. 3B).

Fig. 3.

Tau phosphorylation in TKO mice. (A) Forebrain, hindbrain (brain stem/cerebellum), and spinal cord extracts of three independent WT, KO, TT4, and TKO mice (10 months) subjected to Western blotting with the indicated antibodies. Open arrowheads: transgenic human tau; filled arrowheads: endogenous tau. (B) Western blots with the indicated antibodies of total brain lysates and sarkosyl (in)soluble fractions of these lysates from KO, TT4, TKO, and TPLH mice.

Behavioral Testing of PR61/B’δ KO and TKO Mice.

We examined three- and eight-month-old KO mice, alongside 10-month-old TKO mice for (sensori)motor and cognitive deficits. Motor coordination, evaluated by beam walk and rotarod tests, exhibited poor performance for eight-month-old KO mice in comparison with three-month-old KO and age-matched WT mice (Fig. 4A), correlating well with the age-dependent occurrence of tauopathy. This was further confirmed in TKO mice, where these defects were more than additively affected compared to the single KO or TT4^+/- background (Fig. 4B). Average reaction time on the hot plate was significantly delayed in both KO age groups compared to WT controls (Fig. 4A), indicative of decreased nociception independently of the tau phenotype. This was confirmed in TKO mice, where no additional defects were observed compared to KO mice (Fig. 4B). In adhesive removal tests, only eight-month-old KO mice took more time to remove the adhesive sticker than WT mice (Fig. 4A), but this was not further affected in TKO mice (Fig. 4B). No significant differences between KO or WT animals could be observed in freeze frame and object recognition tests (Fig. 4A).

Fig. 4.

Behavior tests in age- and gender-matched WT, KO, TT4, and TKO mice. (A) In three- or eight-month-old PR61/B’δ KO mice compared to WT (n = 6 for each condition). (B) In 10-month-old TKO mice compared to TT4, KO, and WT (n = 6 for each condition). Statistical significance was evaluated with Student’s t-test (* : p < 0.05; ** : p < 0.01; *** : p < 0.001).

Poor Direct Tau Dephosphorylation by PP2A_T61δ.

Although it is generally believed that PP2A_T55 is the likely in vivo tau phosphatase (11, 12), the simplest explanation for increased AT8/AT180 phosphorylation in the KO would be lack of direct dephosphorylation by PP2A_T61δ. To test this, we exploited tau heat stability to isolate AT8/AT180-positive tau from TPLH mice (13) in the absence of phosphatase inhibitors. Phosphatase assays with de novo purified PP2A_T55 and PP2A_D (18) on this substrate confirmed that PP2A_T55 is by far the better S/T-P site phosphatase (19) (Fig. 5A), even if eightfold more PP2A_D is used (Fig. 5B). In addition, by using carefully equilibrated amounts of PP2A_T55α and PP2A_T61δ retrieved from COS7 cells expressing PR55/Bα and PR61/B’δ GST fusion proteins (20), PP2A_T55α proved at least 15-fold better in dephosphorylating AT8 and AT180 than PP2A_T61δ (Fig. 5 C and D). We confirmed this with recombinant [³²P]-tau in vitro labeled with CDK2/cyclinA at AT8 and AT180 epitopes (21). In this case PP2A_T55α was ±eightfold better than PP2A_T61δ (Fig. 5E) as 10-fold less units of PP2A_T55α dephosphorylate tau with almost equal velocity. Thus, in the presence of PP2A_T55, the absence of PP2A_T61δ in vivo is unlikely to cause tau phosphorylation by lack of direct dephosphorylation.

Fig. 5.

In vitro tau dephosphorylation with various PP2A holoenzymes. (A) Equilibrated amounts of de novo purified PP2A_D and PP2A_T55 were incubated with phospho-tau isolated from TPLH mice for the indicated times. Dephosphorylation was monitored with AT8 and AT180 antibodies. (B) Same experiment with different amounts of PP2A_D (units indicated). (C) Same experiment with PP2A_T55α and PP2A_T61δ isolated from GST-PR55α and GST-PR61δ expressing COS7 cells. Normalized quantifications in D. (E) [³²P]-tau, in vitro labeled by CDK2/cyclinA was incubated with the indicated amounts of PP2A_T61δ or PP2A_T55α. At several time points aliquots were taken and TCA soluble [³²P] measured.

Increased GSK3β Activity in PR61/B’δ KO Mice.

Because many kinases are regulated by PP2A (2), we wondered whether tau phosphorylation might be indirectly the consequence of an effect of PP2A_T61δ on the major in vivo tau AT8/AT180 kinase, GSK3β (8, 22). We observed decreased phosphorylation of the inhibitory GSK3β Ser9 site in KO brain stem/spinal cord, without a change in total GSK3β levels (Fig. 6A). Consequently, GSK3β activity measured with a specifically primed peptide (23) in GSK3β immunoprecipitates (IPs) from brain stem extracts, showed a 30% increase in the KO (Fig. 6B), providing a reasonable explanation for the observed tau hyperphosphorylation. While GSK3β was activated, another potential tau kinase, p42/p44 ERK, did not show any significant activity change in the affected KO brain regions (Fig. S7).

Fig. 6.

GSK3β and CDK5 status in PR61/B’δ KO brainstem and spinal cord. (A) Western blot of brain stem and spinal cord extracts of 18-month-old mice (n = 5) showing decreased GSK3β Ser9 phosphorylation in KO mice. (B) Increased activity in GSK3β IPs from brain stem extracts of 18-month-old KO mice (n = 5) measured with phospho-glycogen synthase peptide 2 substrate. (C) Western blot showing highly reduced levels of p35 in brain stem and spinal cord of three- and 18-month-old KO mice. Positive control: brain extract of p25 transgenic (trg) mice (42). (D) Decreased activity in CDK5 IPs from brain stem extracts of six-month-old KO mice (n = 5) measured with Histone IIA as substrate and roscovitin (10 μM) as specificity control. (E) [³²P]-p35, in vitro labeled by CDK2/cyclinA was incubated for the indicated times with PP2A_T61δ, PP2A_T55α, PP2A_D, or PP1 (all 1 U/ml). Dephosphorylation was monitored by autoradiography following SDS/PAGE and quantified (averages of two independent experiments).

Decreased CDK5 Activity in PR61/B’δ KO Mice.

Because decreased phosphorylation of GSK3β Ser9 was observed in the KO, this site cannot be the direct target of PP2A_T61δ. In the absence of an effect on Akt phosphorylation (Fig. S7), we analyzed activity of another proposed upstream GSK3β Ser9 kinase, the brain-enriched CDK5 (8, 24, 25). While total levels of CDK5 were similar in brain stem and spinal cord of WT and KO mice (Fig. 6C), less than 50% of roscovitin-sensitive histone kinase activity could be measured in CDK5 IPs from KO vs. WT brain stem (Fig. 6D). Surprisingly, this was due to nearly complete absence of the CDK5 activator p35 (as well as its calpain-generated fragment p25) (Fig. 6C). Notably, p35 levels and GSK3β Ser9 phosphorylation were normal in lysates of KO cortex and cerebellum (Fig. S8). This was further confirmed by p35 IHC (Fig. S8). p35 changes are probably due to increased protein degradation, because no quantitative differences in CDK5, p35, or p39 mRNA levels were found by quantitative PCR. Interestingly, previous studies suggested p35 phosphorylation by CDK5 or other kinases regulates its proteasomal degradation, and this is stimulated by okadaic acid (OA), a general PP2A inhibitor (26–28). To test whether PP2A_T61δ might act as a p35 phosphatase, we subjected in vitro CDK2/cyclinA-phosphorylated p35, expressed and purified from bacteria, to dephosphorylation with equal amounts (1 U/ml) of several OA-sensitive phosphatases (Fig. 6E). In contrast to CDK2/cyclinA-phosphorylated tau (Fig. 5), p35 was a good in vitro PP2A_T61δ substrate as this holoenzyme dephosphorylated p35 with at least equal or even better velocity than PP2A_T55, and was significantly better than PP2A_D (Fig. 6E). PP1 hardly dephosphorylated p35 at all (Fig. 6E), confirming the OA effect on p35 stability (27, 28) is due to inhibition of a PP2A phosphatase.

Spatio-Temporal Changes in PR55/Bα and PR61/B’δ Expression in WT and KO Brain.

To further underscore the model emerging from our biochemical data, we wondered whether tauopathy might be correlating with spatio-temporal changes in PP2A subunit expression, focused on PR61/B’δ and PR55/Bα. We found higher absolute levels of PR61/B’δ in WT cortex vs. WT brain stem of young mice and decreased expression upon aging in the cortex but not in brain stem (Fig. S9). In contrast, we found higher PR55/Bα expression in cortex upon aging but not in brain stem (Fig. S9). Finally, PR55/Bα expression in a given brain area and age group did not significantly differ between WT and KO (Fig. S9), suggesting no additional aberrancies due to PR61/B’δ absence.

Discussion

In this study, we present the murine KO of PR61/B’δ, a member of PR61/B’, the largest PP2A B-type subunit family with multiple roles in cellular signaling (2, 29, 30). Despite high embryonic PR61/B’δ expression in WT mice (7), PR61/B’δ KO mice are viable and fertile, with normal placental development and trophoblast differentiation. Because the reported PP2A_T61δ function in dephosphorylation of developmental transcription factor HAND1 specifically is suppressed during trophoblast differentiation (31), this finding was thus not so surprising. Specifically PR61/B’δ also dephosphorylates the Cdc25 Thr138 site to control mitosis (32), but how this might occur in the KO remains unclear given no overt growth abnormalities were observed. Notably, a functional compensation exists for this dephosphorylation as overexpression of Wee-1 (the Cdc25 opposing kinase) was observed in PR61δ KO MEFs (33). Recently, a role was identified specifically for PR61/B’δ in mast cell degranulation (34), but if and how this is affected in the KO remains to be defined.

Given high expression of PR61/B’δ in brain (7), we focused on neural phenotypes in the KO and found evidence for tauopathy, characterized by tau hyperphosphorylation at distinct pathological sites (AT8, AT180, AT100, and AD2) and an altered, pretangle conformation defined by MC1. Remarkably, this tauopathy is spatio-temporally restricted: It localizes in brain stem and the dorsal horn of the cervical spinal cord and becomes evident at six months, subsequently spreading to subiculum, cerebellum, and cortex in 18-month-old mice. The MC1 epitope, highlighting tau misfolding, is one of the earliest pathological alterations in AD tau and absent in normal brain (35). Despite MC1-positivity, we did not observe any NFT pathology, further underscoring that tau hyperphosphorylation and/or misfolding do not necessarily result in NFT formation (36). Our observation of an earlier (six months) increase in HSP90/70 expression in the KO supports the hypothesis that these chaperones may be involved in clearance of MC1-positive tau at that age, thus possibly preventing NFT in older mice, as reported by others (16).

The restricted distribution of tau phosphorylation in the KO, further substantiated by the very same spatial pattern of increased tau phosphorylation in TKO mice, is intriguing given that PR61/B’δ is broadly expressed throughout WT brain (Fig. 1). One might explain this by functional redundancy in other brain regions by other PR61/B’ and possibly PR55/B subunits, most of which are present in brain (7, 37, 38). We did however not notice any significant compensatory changes in expression of several B-type subunits in total brain extracts of young KO mice (three months), nor in PR55/Bα expression in cortex or brainstem of three- or 20-month-old KO mice. Thus potential redundancies in nonaffected KO brain regions might occur by other mechanisms (through other PP2A subunits, other Ser/Thr phosphatases, or even through protein kinases) or are not exemplified by compensatory expression but occur at a functional level. Although it is hardly possible in a biological model as complex as a whole organism to obtain a comprehensive picture of all such possible redundancies, our current and previously published (33) data confirm the existence of general functional compensations. Nonetheless, the pathology in brain stem/spinal cord of aged mice also favors the view of nonredundant functions of PR61/B’δ. Our data indicate preferential rather than exclusive dephosphorylation of substrates by PP2A_T61δ, exemplified by our biochemical assays in which PP2A_T61δ proved only slightly more efficient than PP2A_T55α in dephosphorylating p35 and further underscored by a redundant role of PR61/B’δ in neurotrophin receptor signaling (39). Only if redundant mechanisms entirely or partially fail, are absent, or spatially restricted, a phenotype such as phosphorylated, pretangle tau may appear. Obviously, such observations are difficult to make or reproduce in vitro and thus highlight the importance of in vivo models to understand a protein’s true physiological role.

One of the affected pathways in the KO unexpectedly involves CDK5/p35 signaling. We indeed observed a dramatic decrease in p35, specifically in the affected brain regions, probably via increased proteasomal degradation, although we cannot exclude translation is also affected. Previous studies have suggested phosphorylation regulates p35 degradation and this is stimulated by OA (26–28). Our results would fit into this concept, identifying PP2A_T61δ as a major p35 phosphatase in brain stem/spinal cord: In the absence of PR61/B’δ and presence of normal PR55/Bα levels (Fig. S9), dephosphorylation is hampered resulting in increased p35 degradation. This is sustained by our in vitro assays, showing that in contrast to tau, which is severalfold better dephosphorylated by PP2A_T55, p35 is a relatively much better PP2A_T61δ substrate. Although CDK5 has been repeatedly reported as a tau kinase, CDK5 deregulation resulting in NFT in vivo is rather linked to pathological generation of p25, a calpain-dependent proteolytic p35 fragment (40) that is absent in the affected KO brain regions. Decreased CDK5 activity in the KO is in any case inconsistent with CDK5 directly phosphorylating the pathological tau epitopes but concurs with transgenic mice in which overexpression of CDK5 and/or p25 does not lead to AT8/AT180 hyperphosphorylation (41, 42). Similarly, CDK5 KO mice die perinatally with increased phosphorylation of neurofilaments in brain stem (43). These observations add to the controversy whether CDK5, activated by p35, can act as a physiological tau kinase in vivo (8) and rather support CDK5 is a “negative tau kinase” indirectly affecting tau phosphorylation through control of GSK3β (8, 24, 25, 41). Our results would sustain this, because we observed decreased GSK3β Ser9 phosphorylation and subsequent kinase activation, reminiscent of the overt GSK3β activation in p35 KO mice (44). As GSK3β is an established tau AT8/AT180 kinase in vivo (8, 22, 41, 45, 46) this elevated GSK3β activity is the likely cause of tau hyperphosphorylation in the KO. Exactly how CDK5 promotes GSK3β Ser9 phosphorylation remains elusive and may be achieved through direct as well as indirect mechanisms—for instance by controlling PP1 activity (24, 47). By no means, however, do our data support that PP2A_T61δ is a GSK3β Ser9 phosphatase, opposing recent findings in a genomic poorly characterized gene trap model (48). Whether PR61/B’δ-mediated control of p35 stability has wider implications on other CDK5 substrates, such as the predicted PP2A_T61δ substrate DARPP-32 (49, 50) awaits further exploration. Likewise, at least one behavioral defect—the increased latency on a hot plate—might be related to the role of CDK5 in nociception (51). Our other behavioral studies provided clear support for impairment of some (sensori)motor and coordination functions in the KO (rotarod/beamwalk). In these cases a causal relationship with tauopathy is suggested by the age-dependency of the defects in the KO and the more than additive defects in TKO mice. Such a relationship would also imply that tauopathy can result in motoric defects in the absence of NFT or neurodegeneration—for instance by affecting proper neuronal function. Finally, cognitive functions in the KO appeared normal, consistent with absence of tauopathy in areas such as hippocampus.

Several reports have highlighted PP2A impairment is clinically relevant in AD. In addition to decreased enzyme activity, mRNA levels of several PP2A subunits are significantly decreased in AD brain (52, 53). In transgenic mice expressing a dominant-negative PP2A_C mutant (L199P), tau is hyperphosphorylated at AT8 and Ser422 (54). AT8 phosphorylation also occurred in mice expressing PP2A_C L309A (55), specifically deficient in PR55/B recruitment (1). Although it is generally believed and confirmed here that PP2A_T55 is the major tau phosphatase (9–12, 56) and PP2A_T55α protein levels are reduced in AD brain (53), PR61/B’ expression can be equally well affected (52), so far with unclear consequences. Based on the present study it is tempting to speculate that specifically PR61/B’δ may be involved early in the disease etiology, not necessarily by a direct mechanism, but by maintaining a proper kinase-phosphatase balance.

Materials and Methods

Protocols for the generation of PR61/B’δ KO and TKO mice and for behavior tests can be found in SI Materials and Methods. Tissue extracts and sarkosyl (in)soluble tau were prepared as described (13). Procedures for brain/spinal cord collection, IHC and CongoRed/X34 staining were as described (13, 42). Primary antibodies can be found in Table S1. PC12 IF and protocols for kinase and phosphatase assays are given in SI Materials and Methods.