The Spinocerebellar Ataxia 12 Gene Product and Protein Phosphatase 2A Regulatory Subunit Bβ2 Antagonizes Neuronal Survival by Promoting Mitochondrial Fission*

The neurodegenerative disorder spinocerebellar ataxia 12 (SCA12) is caused by CAG repeat expansion in the non-coding region of the PPP2R2B gene. PPP2R2B encodes Bβ1 and Bβ2, alternatively spliced and neuron-specific regulatory subunits of the protein phosphatase 2A (PP2A) holoenzyme. We show here that in PC12 cells and hippocampal neurons, cell stressors induced a rapid translocation of PP2A/Bβ2 to mitochondria to promote apoptosis. Conversely, silencing of PP2A/Bβ2 protected hippocampal neurons against free radical-mediated, excitotoxic, and ischemic insults. Evidence is accumulating that the mitochondrial fission/fusion equilibrium is an important determinant of cell survival. Accordingly, we found that Bβ2 expression induces mitochondrial fragmentation, whereas Bβ2 silencing or inhibition resulted in mitochondrial elongation. Based on epistasis experiments involving Bcl2 and core components of the mitochondrial fission machinery (Fis1 and dynamin-related protein 1), mitochondrial fragmentation occurs upstream of apoptosis and is both necessary and sufficient for hippocampal neuron death. Our data provide the first example of a proapoptotic phosphatase that predisposes to neuronal death by promoting mitochondrial division and point to a possible imbalance of the mitochondrial morphogenetic equilibrium in the pathogenesis of SCA12.

Mitochondrial morphology and assembly of mitochondria into a contiguous network is a consequence of opposing fission and fusion events (1)(2)(3). In neurons, physiological levels of mitochondrial fission are necessary for axonal and dendritic transport of mitochondria and the proper development and function of synapses (4,5). However, mitochondrial fragmentation is also an integral aspect of apoptosis because enhancing and inhibiting fission can promote and delay apoptosis, respectively (6,7). Mitochondrial fission and fusion are separate processes catalyzed by large GTPases that were initially identified in yeast (8). The principal fission enzyme is dynamin-related protein 1 (Drp1), 5 which, in analogy to the endocytosis motor dynamin, is thought to utilize GTP hydrolysis to mechanically constrict and sever mitochondria. Drp1 is recruited from the cytosol to the outer mitochondrial membrane (OMM) by a multiprotein complex that includes Fis1. In mammals, outer mitochondrial membrane fusion is carried out by two transmembrane GTPases, mitofusin 1 and 2 (Mfn1/2), which act in concert with the inner membrane fusion enzyme optic atrophy 1 (Opa1) (8).
The fundamental importance of a healthy mitochondrial fission/fusion balance in neurons is documented by the discovery that two common neurodegenerative disorders, Charcot-Marie-Tooth disease type 2A and dominant optic atrophy, are caused by mutations in mitochondrial fusion GTPases, Mfn2 and Opa1, respectively (9 -11). Similarly, a dominant-inactivating mutation in Drp1 was recently linked to microcephaly and other neurological birth defects (12). Despite the clear relevance of mitochondrial morphogenesis to neuronal survival (6,(13)(14)(15), we know very little about the signal transduction pathways that impinge on mitochondrial remodeling processes and their physiological and pathophysiological sequelae.
Protein phosphatase 2A (PP2A) is a major, evolutionarily conserved family of Ser/Thr phosphatases with many essential functions in the cell (16). The predominant form of PP2A is a heterotrimer of a catalytic C and scaffolding A and a variable regulatory or B subunit. Regulatory subunits define substrate specificity and subcellular localization of the PP2A holoenzyme and are often expressed in a tissue-and cell type-specific pattern. As one of the 12 mammalian PP2A regulatory subunit genes, B␤ (PPP2R2B) gives rise to several neuron-specific splice variants (17,18). Indicating a role for PP2A/B␤ in neuronal survival, a CAG repeat expansion in a non-coding region of PPP2R2B was shown to be responsible for spinocerebellar ataxia type 12, SCA12 (19). SCA12 is a relatively rare, late-onset neurodegenerative disorder characterized by diffuse cerebral and cerebellar atrophy (20). It is not known whether and how the CAG repeat expansion affects expression of the different B␤ splice variants in humans, and animal models of SCA12 have yet to be established.
We previously reported that the differentially spliced N terminus of B␤2 recruits PP2A to the OMM by a transient and reversible interaction with receptor subunits of the translocase of the outer membrane (TOM) complex. Furthermore, overexpression of B␤2 was shown to promote apoptosis in growth factor-deprived neuronal PC12 cells (18,21). Here, we demonstrate that B␤2 redistributes from the cytosol to mitochondria in dying PC12 cells and hippocampal neurons and that apoptosis induction by B␤2 requires recruitment of the PP2A holoenzyme to the OMM. RNA interference-mediated knockdown of B␤2 in hippocampal cultures protects neurons from excitotoxic, ischemic, and metabolic insults. Multiple lines of gainand loss-of-function evidence indicate that PP2A/B␤2 antagonizes survival by driving Drp1-and Fis1-dependent mitochondrial fragmentation. The survival and morphogenetic activities of PP2A/B␤2 can be dissociated by Bcl2 overexpression, indicating that mitochondrial restructuring occurs upstream of apoptosis. Our results identify an outer mitochondrial PP2A holoenzyme as a critical regulator of the mitochondrial morphogenetic equilibrium, which determines the susceptibility of neurons to diverse injuries.
Antibodies-A mouse monoclonal antibody directed against B␤2 was generated at the University of Iowa Hybridoma Facility, using the N-terminal peptide KCFSRYLPYIFRPPNTILSS coupled to maleimide-activated keyhole limpet hemocyanin (Pierce) as an antigen. This antibody specifically recognized ectopically expressed B␤2, but not B␤1, by immunoblotting. As further confirmation of specificity, immunohistochemical labeling of brain sections with the B␤2 antibody was nearly abolished in homozygous B␤2 knock-out mice. 6 Other antibodies were obtained commercially: sheep anti-cytochrome c (Sigma), rabbit anti-GFP (AbCam), mouse anti-neurofilament (2H3, Developmental Studies Hybridoma Bank, University of Iowa), and rabbit anti-␤-galactosidase (Invitrogen).
Immunofluorescence Staining and Quantification-Hippocampal neurons from E18 rat embryos (25) were cultured in Neurobasal medium with B27 supplement (Invitrogen). At 18 -21 days in vitro, cultures were challenged as indicated, fixed in 4% paraformaldehyde after 2 h, and processed for B␤2 and cytochrome c immunofluorescence. The cytochrome c antibody was visualized by a Alexa Fluor 488-coupled secondary antibody (Invitrogen), whereas B␤2 was detected with the Cy3 tyramide signal amplification kit (PerkinElmer Life Sciences) according to the manufacturer's instructions. Digital images of randomly selected neurons were captured at constant camera settings using the 40ϫ objective of a Leica epifluorescence microscope. Clustering of B␤2 in dendrites was quantified by image analysis using a macro custom-written for the ImageJ program (National Institutes of Health). To this end, images were background-subtracted using the rolling ball algorithm, and an autosegmentation mask was applied outlining regions of above threshold pixel intensity in the dendritic field. Average B␤2 immunofluorescence (8-bit pixel intensity) in outlined dendritic clusters was compiled for 50 -100 neurons per condition. For confocal microscopy of endogenous B␤2 in the soma of hippocampal neurons, cultures were infected with lentivirus expressing mitochondria-targeted GFP (matrix import sequence from cytochrome oxidase subunit VIII) and challenged 5 days later with 2.5 M rotenone or DMSO vehicle. After 2 h, cultures were paraformaldehyde-fixed, subjected to antigen retrieval (10 mM Tris, 1 mM EDTA, 0.05% Tween 20, pH 9.0, for 30 min at 80°C), and then immunofluorescently labeled for B␤2 and GFP as described above. Confocal image stacks (3-m interval) were acquired with an Olympus FluoView 3000 microscope, and colocalization of B␤2 with mitochondria was assessed by blinded comparison with a set of reference images.
Neuronal Survival Assays-Hippocampal cultures (25) were transduced with lentivirus or transfected using Lipofectamine 2000 (0.15%, 2 g/ml DNA) at 10 -21 days in vitro. For survival assays based on counting transfected neurons or apoptotic nuclei, GFP-or ␤-galactosidase-expressing plasmids were cotransfected. For oxygen-glucose deprivation assays, a plasmid expressing luciferase driven by the neuron-specific calcitonin gene-related peptide promoter was included (20% of total plasmid mass). After 3-5 days, cultures were challenged with glutamate (250 M, 20 min), rotenone (400 nM continuous), or oxygen-glucose deprivation (25 min). Glutamate-and rotenone-treated cultures (and their vehicle controls) were fixed with 3.7% paraformaldehyde and processed for immunofluorescence staining for the transfection marker (GFP, ␤-galactosidase) and neurofilament protein as a neuronal marker. To score survival, transfection marker-positive neurons with intact processes were counted in quadruplicate wells of a 24-well plate. Apoptosis was quantified as the percentage of transfected neurons with condensed, irregular, or fragmented nuclei (labeled with 1 g/ml Hoechst 33342). All survival assays based on counting neurons were performed blind to the experimental conditions.
For oxygen-glucose deprivation assays, medium was exchanged with custom Neurobasal medium without glucose (formula 05-0128DJ, Invitrogen), and cultures were placed for 25 min in a Billups-Rothenberg (Del Mar, CA) modular incubator chamber, which was flushed with 95% N 2 and 5% CO 2 . After adding back glucose-containing Neurobasal medium, cultures were returned to the standard incubator. Control cultures were mock-treated under normoglycemic/normoxic conditions. 72 h later, neuronal survival was determined by luciferase assays.
Mitochondrial Morphology Assay-PC12 cells and hippocampal neurons cultured on, respectively, collagen-and poly-D-lysine-coated, chambered Number 1 cover glasses (20-mm 2 chamber, Nalge Nunc), were infected with lentivirus or transfected using Lipofectamine 2000 as above. 24 -96 h after transfection, cells were incubated with 100 nM tetramethylrhodamine methyl ester (TMRM, Invitrogen) for 30 min at 37°C to visualize mitochondria, and live cell images through the midplane of the soma were captured using a Zeiss LSM 510 laser-scanning confocal microscope. Mitochon-drial morphology was scored by a reference image-based method. Coded confocal images were assigned scores from 0 to 4 by comparison with a set of reference images with increasing degrees of mitochondrial elongation. Nontransfected or non-transduced cells were analyzed in some experiments, yielding mitochondrial morphology scores that were statistically indistinguishable from cells expressing control shRNA or control proteins (e.g. outer mitochondrial GFP).
Statistical Analysis-Data were analyzed by Student's t test (twotailed) for single comparisons, and by one-way analysis of variance followed by pairwise Bonferroni post hoc tests for multiple comparisons.

PP2A/B␤2 Translocates to Mitochondria to Promote Neuronal Death-
We investigated a possible redistribution of B␤2-GFP in healthy versus stressed PC12 cells. Under basal conditions, B␤2 localized to both cytosolic and mitochondrial compartments (the latter visualized by MitoTracker or TMRM) as reported previously (18). In contrast, B␤1-GFP localization was exclusively cytosolic, excluding areas of high mitochondrial density (18). When cells were incubated with the glycosylation inhibitor tunicamycin (1.4 M), which induces apoptosis through the endoplasmic reticulum stress pathway, B␤2 assumed an exclusively mitochondrial localization (Fig. 1A). Tunicamycin-dependent mitochondrial redistribution of B␤2 was quantified at 24-and 48-h time points (Fig. 1B). Rotenone is a pesticide that blocks complex I of the electron transport chain and that has been used as a chemical model of Parkinson disease (26). Similar to tunicamycin, rotenone (10 M) induced a quantitative translocation of B␤2 from the cytosolic to the mitochondrial compartment, which correlated in time and on a cell-by-cell basis with nuclear condensation and fragmentation, hallmarks of apoptosis (Fig. 1C).
We extended these results to primary hippocampal neurons transduced with lentivirus expressing B␤1-or B␤2-GFP. As in PC12 cells, B␤1-GFP was excluded from mitochondria (labeled with the potential-sensitive dye TMRM). Confocal imaging of B␤2-GFP-expressing neurons revealed a modest degree of colocalization with mitochondria (Fig. 1D). Treatment of hippocampal cultures with 2 M rotenone resulted in a rapid relocalization of B␤2, but not B␤1, from the cytosolic to the mitochondrial compartment (Fig. 1D). As in PC12 cells,  DECEMBER 26, 2008 • VOLUME 283 • NUMBER 52 mitochondrial translocation of B␤2 was tightly correlated with the appearance of apoptotic nuclei and occurred concomitantly (Fig. 1E). We also observed increased association of B␤2-GFP with mitochondria of PC12 cells following growth factor deprivation (data not shown).

PP2A/B␤2 Promotes Mitochondrial Fragmentation
To investigate the localization of endogenous B␤2, a monoclonal antibody was generated against a unique N-terminal sequence. This antibody is specific for B␤2 as it labels various neuronal populations in brain sections of wild-type, but not homozygous B␤2 null mice. 6 Hippocampal cultures were challenged with glutamate (200 M, 15 min), rotenone (2.5 M, 2 h), or the classical apoptosis inducer staurosporine (0.5 M, 2 h), fixed after 2 h, and immunofluorescently labeled for endogenous B␤2 and mitochondrial markers. We used epifluorescence microcopy and computer-aided image analysis to quantify B␤2 translocation to cytochrome c-positive areas in dendrites and detected significant increases with all insults (Fig. 2, A and C). To assess stress-induced association of B␤2 with mitochondria in the soma of hippocampal neurons, cultures infected with lentivirus expressing mitochondrial matrix-targeted GFP were subjected to a 2-h DMSO or rotenone treatment, immunolabeled for B␤2, and analyzed by confocal microscopy. Paralleling the behavior of GFP-tagged B␤2 (Fig. 1, D and E), rotenoneinduced mitochondrial swelling and fragmentation was accompanied by a transition of endogenous B␤2 from a cytosolic to a largely mitochondrial staining pattern (Fig. 2, B and D). These data indicate that mitochondrial redistribution of this neuronspecific PP2A regulatory subunit may be a general consequence of neuronal injury.

Mitochondrial Targeting of the PP2A Holoenzyme by B␤2 Is
Neurotoxic-When compared with non-transduced neurons in the same culture, B␤2-GFP-overexpressing hippocampal neurons displayed increased vulnerability to rotenone (Fig. 1E). Therefore, translocation of B␤2 to mitochondria appears to be a positive feedback mechanism that accelerates neuronal death. In support of this conclusion, long term (Ն5 days) expression of B␤2-, but not B␤1-GFP, in hippocampal neurons resulted in massive apoptosis (Fig. 3A). We previously identified critical salt bridge interactions between Arg-168/Arg-169 of B␤2 and Glu-100/Glu-101 of the scaffolding A subunit and characterized in detail the determinants for outer mitochondrial localization of B␤2 (18,21,27). In further overexpression experiments, we found that the proapoptotic activity of B␤2 requires recruitment of the PP2A holoenzyme to the mitochondrial surface, because mutations that disrupt association with mitochondria (R6A) or with the scaffolding and catalytic subunit (RR168EE) blocked the ability of B␤2 to kill hippocampal neurons (Fig. 3A).
PP2A/B␤2 Silencing Is Neuroprotective-We employed RNA interference to investigate whether endogenous B␤2 is a negative regulator of neuronal survival. Two shRNAs targeting the splicespecific 5Ј-untranslated region and N-terminal coding sequence effectively silenced a B␤2-luciferase fusion protein in hippocampal cultures and stably expressed B␤2 in PC12 cells (supplemental Fig.  1). B␤2 knockdown conferred robust neuroprotection in hippocampal cultures exposed to a brief excitotoxic glutamate treatment as determined by counting the number of transfected neurons with intact cell bodies and processes (Fig. 3, B and C). The representative images in Fig. 3B illustrate the enhanced gluta- mate resistance of B␤2 shRNA-expressing neurons when compared with their untransfected neighbors. Secondly, in an in vitro model of ischemia (oxygen/glucose deprivation), B␤2 knockdown attenuated hippocampal neuron loss, which in these experiments was measured by luciferase expression from a neuron-specific promoter (Fig. 3D). Lastly, B␤2 silencing by two shRNAs increased the number of surviving neurons and decreased apoptosis following inhibition of mitochondrial electron transport with rotenone (Fig. 3, C and E). In aggregate, these results implicate the SCA12 disease-associated PP2A/ B␤2 regulatory subunit as a neuron-specific component of the cell death machinery.
PP2A/B␤2 Drives Mitochondrial Fission-Mitochondrial fragmentation is an early, obligatory step in the intrinsic, mitochondria-dependent apoptosis program (6,7,28). In hippocampal cultures stained with the mitochondrial voltage-sensitive dye TMRM, we observed that neurons expressing B␤2-GFP showed more punctate mitochondria than B␤1-expressing neurons (Fig. 4A). Similar results were obtained in dividing and differentiated PC12 cells, as well as in dissociated dorsal root ganglion neurons (supplemental Fig. Fig. 2, B-D). To quantify mitochondrial shape changes, coded confocal images through the midplane of the neuronal soma were compared with a set of reference images with increasing degrees of mitochondrial length and interconnectivity (supplemental Fig. 2A, score 0 -4).
Although expression of wild-type B␤2 resulted in significant mitochondrial fragmentation, the monomeric B␤2 mutant (RR168EE (18)) had an apparent dominantnegative effect, elongating and often clustering mitochondria near the nucleus (Fig. 4, A and B). RNA interference-mediated silencing of endogenous B␤2 with two different shRNAs resulted in robust mitochondrial elongation to an extent similar to dominant-negative inhibition of the PP2A-targeting subunit (Fig. 4, A and B).
To support a role for PP2A activity in mitochondrial morphogenesis, we expressed Plasmodium falciparum aspartate-rich protein (PfARP) in hippocampal neurons. PfARP is a predominantly cytosolic member of the I2 PP2A family of PP2A catalytic subunit inhibitors (22). Expression of an untargeted PfARP-GFP fusion protein had no effect on mitochondrial morphology. However, redirecting the PP2A inhibitor to mitochondria by fusing the OMM anchor sequence of the mitochondrial import receptor MAS70p to its N terminus generated longer mitochondria (Fig. 4B). These results suggest that OMM targeting of PP2A activity via the neuron-specific B␤2 regulatory subunit promotes cell death by shifting the mitochondrial fission/fusion balance toward fission.
Mitochondrial Remodeling Is Critical for Neuronal Survival Regulation-Mitochondrial fragmentation induced by B␤2 expression generally preceded loss of mitochondrial membrane potential (TMRM staining) and the appearance of apoptotic nuclei. Our results are therefore consistent with a model in which death is epistatic to fission. As a first test, we showed that rotenone-induced apoptosis, but not mitochondrial fragmentation, can be prevented by expressing the anti-apoptotic Bcl2 protein in hippocampal neurons (Fig. 5). Transfection of the fission GTPase Drp1 exacerbated both basal and rotenone-induced apoptosis. In agreement with a previous report (13), expression of a dominant-negative Drp1 mutant (K38A) or shRNAmediated silencing of the adaptor protein Fis1 (supplemental Fig. 1A), which recruits Drp1 to the OMM, not only prevented mitochondrial fragmentation but also apoptosis induced by rotenone (Fig. 5). Thus, Fis1/Drp1-mediated mitochondrial fission is necessary and sufficient for at least some forms of neuronal cell death.
We next investigated the involvement of mitochondrial fission in neuronal apoptosis induced by expression of B␤2. Bcl2 coexpression completely blocked cell death associated with recruiting PP2A to mitochondria but did not restore normal mito- chondrial morphology (Fig. 6). Therefore, PP2A-mediated mitochondrial restructuring is upstream of and distinct from the disintegration of various organelles that accompanies cell death.
To address whether mitochondrial morphogenetic events are required for the death-promoting activity PP2A/B␤2, we expressed B␤2 together with dominant-negative Drp1-or Fis1directed shRNA. Interfering with Fis1 and Drp1 not only reinstated normal mitochondrial morphology (Fig. 6A) but also at least partially restored viability of hippocampal neurons challenged by B␤2 overexpression (Fig. 6B).

DISCUSSION
We have identified the SCA12 gene product B␤2 as a neuronspecific, OMM-localized PP2A regulatory subunit that regulates neuronal survival through the mitochondrial fission/fusion balance. RNA interference-mediated silencing of PP2A/ B␤2 was shown to confer neuroprotection against excitotoxic, ischemic, and free radical-mediated injury. Because OMM targeting is necessary for the proapoptotic activity of B␤2, agents that interfere with the injury-induced mitochondrial translocation of this neuron-specific PP2A regulatory subunit should also be neuroprotective.
The CAG repeat expansion in SCA12 maps close to the transcript initiation site of the B␤1 splice form (20), which lacks the mitochondrial localization and proapoptotic activity of B␤2. Because B␤2-specific exons are located upstream of the B␤1 transcript start site, the B␤2 pre-mRNA necessarily includes the CAG repeat. Although as yet untested, it is conceivable that the repeat expansion increases B␤2 mRNA levels, resulting in late-onset neurodegeneration. Ultimately, however, animal models are required to address whether a disturbance of the mitochondrial fission/fusion equilibrium is the primary cause of neuron loss in SCA12.
Tilting the kinase/phosphatase balance toward dephosphorylation by recruiting PP2A to the OMM (via wild-type B␤2 expression) resulted in smaller mitochondria and compromised neuronal viability, whereas silencing PP2A/〉␤2 had the opposite effect. Neuronal death induced by favoring outer mitochondrial dephosphorylations could be mitigated by expression of Bcl2, dominant-negative Drp1, and Fis1 knockdown. Importantly, Bcl2 did not prevent mitochondrial fragmentation, supporting a linear model in which PP2A-mediated mitochondrial remodeling is epistatic to apoptosis (Fig. 6C).
In agreement with similar results from other laboratories (13)(14)(15)29), we found that inhibition of Fis1 and Drp1 also potently attenuates neurotoxicity of the electron transport inhibitor rotenone. Although the contribution of mitochondrial shape changes to apoptosis in immortalized cell lines is currently a matter of controversy (30,31), Fis1/Drp1-dependent mitochondrial fragmentation appears to be a critical checkpoint in the demise of primary neurons induced by either proapoptotic proteins or environmental toxins.
The role of PP2A in programmed cell death is well established (3,32,33), and PP2A can dephosphorylate both proapoptotic and antiapoptotic Bcl2-family proteins (34 -36). Here, we provide evidence for a novel mechanism of cell death regu-  lation by PP2A (Fig. 6C). We hypothesize that in healthy neurons, as yet to be identified kinases drive mitochondrial fusion and/or inhibit fission by phosphorylation of the enzymes involved. Upon cell stress (e.g. glutamate, rotenone, staurosporine), the PP2A/B␤2 holoenzyme translocates from the cytosol to the OMM via an interaction of a cryptic mitochondrial import sequence in the alternatively spliced N terminus of B␤2 and receptor components of the TOM import complex (21). Outer mitochondrial PP2A, perhaps in conjunction with kinase inactivation, proceeds to dephosphorylate mitochondria-shaping proteins, thus inhibiting fusion and/or stimulating fission activities. Excessive mitochondrial fragmentation could lead to cell death either by fostering the release of apoptogenic proteins from the mitochondrial intermembrane space or by altering mitochondrial reactive oxygen species and Ca 2ϩ homeostasis (1-3, 6, 28).
Drp1 was recently shown to be phosphorylated at two sites by cyclin-dependent kinase 1, protein kinase A, and calcium/ calmodulin-dependent kinase I, with somewhat controversial effects on the GTPase and mitochondria-severing activities of the enzyme (37)(38)(39)(40)(41). Rather than PP2A, however, the calcium/ calmodulin-dependent phosphatase PP2B/calcineurin has been implicated in the dephosphorylation and activation of Drp1 (39,40). Thus, further studies are needed to identify the relevant PP2A substrates and dephosphorylation sites among mitochondria restructuring enzymes and their known and unknown modulators (8).
Although B␤2 is a neuron-specific PP2A regulatory subunit, other mitochondrial PP2A holoenzymes have been reported in non-neuronal cells (34). It is therefore likely that similar survival signaling mechanisms operate outside of the nervous system. In summary, our study points to the mitochondrial fission/ fusion equilibrium as a pivotal determinant for neuronal life and death decisions, identifies the phosphatase PP2A/B␤2 as an important modulator, and suggests dysregulation of mitochondrial morphogenesis as a plausible pathogenic mechanism in SCA12.