N-terminal phosphorylation of PP2A/Bβ2 regulates translocation to mitochondria, Drp1 dephosphorylation, and neuronal survival

Summary

The neuron-specific Bβ2 regulatory subunit of protein phosphatase 2A (PP2A), a product of the spinocerebellar ataxia type 12 disease gene PPP2R2B, recruits heterotrimeric PP2A to the outer mitochondrial membrane (OMM) through its N-terminal mitochondrial targeting sequence. OMM-localized PP2A/Bβ2 induces mitochondrial fragmentation, thereby increasing susceptibility to neuronal insults. Here, we report that PP2A/Bβ2 activates the mitochondrial fission enzyme dynamin-related protein 1 (Drp1) by dephosphorylating Ser656, a highly conserved inhibitory phosphorylation site targeted by the neuroprotective PKA/AKAP1 kinase complex. We further show that translocation of PP2A/Bβ2 to mitochondria is regulated by phosphorylation of Bβ2 at three N-terminal Ser residues. Phosphomimetic substitution of Ser20-22 renders Bβ2 cytosolic, blocks Drp1 dephosphorylation and mitochondrial fragmentation, and abolishes the ability of Bβ2 overexpression to induce apoptosis in cultured hippocampal neurons. Ala substitution of Ser20-22 to prevent phosphorylation has the opposite effect, promoting association of Bβ2 with mitochondria, Drp1 dephosphorylation, mitochondrial fission, and neuronal death. OMM translocation of Bβ2 can be attenuated by mutation of residues in close proximity to the catalytic site, but only if Ser20-22 are available for phosphorylation, suggesting that PP2A/Bβ2 autodephosphorylation is necessary for OMM association, likely by uncovering the net positive charge of the mitochondrial targeting sequence. These results reveal another layer of complexity in the regulation of the mitochondrial fission/fusion equilibrium and its physiological and pathophysiological consequences in the nervous system.

Introduction

Mitochondria play a vital role in nearly every facet of eukaryotic life, and many mitochondrial functions including ATP synthesis and calcium buffering can be influenced by their size and shape. Mitochondrial shape is a product of the opposing processes of fission and fusion, which are carried out by a group of large GTPases of the dynamin superfamily [1]. Dynamin-related protein 1 (Drp1) is an evolutionary ancient motor protein that catalyzes the mitochondrial fission reaction. In analogy to dynamin, Drp1 is thought to form spirals around mitochondria, which upon GTP hydrolysis constrict to sever the organelle [2]. Drp1-mediated mitochondrial fission is essential in both mouse and man and Drp1 activity is tightly regulated by several posttranslational mechanisms [3-6]. A conserved and well-established regulatory site is serine 656, which upon phosphorylation by protein kinase A (PKA) decreases Drp1 activity to elongate mitochondria by unopposed fusion [7, 8]. We have recently reported that targeting of PKA to the OMM by A kinase anchoring protein 1 (AKAP1) confers neuroprotection that depends on Drp1 S656 phosphorylation and elongation of mitochondria [9].

In opposition to PKA, the ubiquitously expressed calcium-dependent phosphatase calcineurin (CaN or PP2B) has been shown to activate Drp1 via dephosphorylation of S656 [8, 10]. However, our previous work also implicated protein phosphatase 2A (PP2A) as an important regulator of mitochondrial shape and its physiological and pathophysiological sequelae [11, 12]. Heterotrimeric PP2A consists of a dimer containing an A scaffolding subunit bound to a C catalytic subunit plus a variable regulatory B subunit. Encoded by 15 genes in mammals, B regulatory subunits direct localization and substrate specificity of the heterotrimeric complex [13, 14]. The Bβ gene (PPP2R2B) encodes two neuron-specific PP2A regulatory subunits, Bβ1 and Bβ2, and is associated with a debilitating neurodegenerative disease, spinocerebellar ataxia type 12 (SCA12). SCA12 is caused by a CAG repeat expansion within the promoter that drives expression of the Bβ1 variant [15]. Bβ1 and Bβ2 differ only in their first exons encoding 21 and 24 amino acids, respectively. The Bβ1 splice variant directs the heterotrimeric PP2A complex to the cytosol, while Bβ2 localizes PP2A/Bβ2 to both cytosol and mitochondria [16]. Bβ2 is recruited to mitochondria via its alternatively spliced N-terminus, with conserved basic and hydrophobic residues playing critical roles. Bβ2_1-24 acts as a cryptic mitochondrial import signal, since these residues promote matrix import and cleavage by signal peptidases when fused to GFP. However, full length Bβ2 is retained at the OMM via low affinity interactions with receptor components of the translocase of outer membrane (TOM) complex, because Bβ2’s C-terminal β-propeller domain resists the partial unfolding step required for transfer through the TOM40 channel [17]. Knockdown of Bβ2 in hippocampal neurons leads to elongation of mitochondria and confers neuroprotection in models of excitotoxic/ischemic injury. Conversely, overexpression of wild-type, but not OMM targeting-defective PP2A/Bβ2 results in Drp1-dependent mitochondrial fission and increases basal rates of neuronal apoptosis[11, 16]. PP2A/Bβ2 also plays a role in dendrite and synapse development, as knockdown of Bβ2 increases dendritic branch complexity, but decreases synapse number in cultured hippocampal neurons. Bβ2 overexpression promotes synaptogenesis, but this effect could be blocked with pseudophosphorylated Drp1 (S656D mutant) [12]. However, whether PP2A/Bβ2 can dephosphorylate Drp1 was left unanswered.

We have previously shown that various cell stressors (rotenone, glutamate, growth factor withdrawal) cause a dramatic redistribution of Bβ2 from the cytosol to the OMM in PC12 cells and hippocampal neurons [11]. Here, we report on the mechanism of this translocation. We show that Bβ2 localization is determined by the phosphorylation state of its N-terminus, with phosphorylation of Ser20-22 causing sequestration in the cytosol and autodephosphorylation promoting translocation of PP2A/Bβ2 to the OMM. We also provide evidence that Bβ2 N-terminal phosphorylation is neuroprotective, since it inhibits Drp1 dephosphorylation/activation, mitochondrial fragmentation, and neuronal death.

Results

Bβ2 is phosphorylated on N-terminal serines

Mitochondrial import sequences generally carry a net positive charge important for translocation into the negatively charged mitochondrial matrix by an electrophoretic mechanism [18]. The cryptic import sequence of Bβ2 (aa 1-24) includes one threonine, 2 tyrosines, and 4 serines, phosphorylation of which could neutralize the +3 charge of the N-terminus. In initial experiments, we fused Bβ2 N-terminal sequences of increasing length to the N-terminus of GFP and isolated fusion proteins from transfected COS cells by immunoprecipitation. Bβ2-GFP fusion proteins were used as substrates in in vitro phosphorylation reactions with calcium/calmodulin-dependent kinase II and [³³P-γ]ATP under conditions that favor promiscuous phosphorylation of nonconsensus sites. These experiments revealed that residues between position 20 and 26 can be phosphorylated (Fig. 1A). To investigate whether Bβ2 is phosphorylated in intact cells, we immunoprecipitated the FLAG-tagged regulatory subunit from transiently transfected COS1 after metabolic labeling with ³²PO₄^3-. Bβ2 incorporated about twice as much ³²P as the cytosolic N-terminal splice variant Bβ1, indicating that Bβ2 is phosphorylated at residues in the differentially spliced N-terminal tail and the common C-terminal β-propeller domain (Fig. 1B). Additional metabolic labeling experiments with mutant Bβ2 carrying alanines in place of serines 20-22 confirmed phosphorylation of N-terminal residues (Fig. 1C). For further evidence, we examined phosphorylation of the isolated Bβ2 N-terminus (Bβ2_1-35-GFP) in intact cells. The wild-type N-terminus was appreciably phosphorylated, but mutation of serines 20-22 eliminated almost all ³²P incorporation (Fig. 1D). ³²P labeling of Bβ2_1-35-GFP could be detected without inhibiting protein phosphatases. In contrast, ³²P incorporation into full-length Bβ2 (or Bβ1), which incorporates into the PP2A heterotrimer, required treatment with the cell-permeant PP1/PP2A inhibitor calyculin A (25 nM, 30 min) prior to cell lysis and immunoprecipitation. These results indicate that Bβ2 is phosphorylated on one or more of three N-terminal serines, but that these phosphates turn over rapidly, presumably due to autodephosphorylation by the PP2A holoenzyme.

Fig. 1.

Bβ2 can be phosphorylated on N-terminal residues in vitro and in intact cells. (A) N-terminal fragments of Bβ2 fused to GFP were in vitro phosphorylated with purified CaMKII and [γ-³³P] ATP. Major ³³P incorporation occurs between residues 20 and 26 (drop from 92% phosphorylation of Bβ2_1-26 to 18% of Bβ2_1-19). Percent phosphorylation (% phos) was determined by densitometry as the ratio of ³³P to protein signals (Ponceau S total protein stain) relative to Bβ2_1-35 (100%). (B-D) Full-length Bβ1, Bβ2, and Bβ2 SSS20AAA (FLAG-tagged in (B), V5-tagged in (C)) or Bβ2_1-35-FLAG-GFP ((D), wild-type and SSS20AAA) was metabolically labeled with ortho-³²P phosphate in transfected COS1 cells and immunoprecipitated. Cells expressing full-length regulatory subunits (B-C) were treated with the phosphatase inhibitor calyculin A (25 nM, 30 min) prior to lysis. % phos is the ratio of ³²P to immunoblot signals relative to wild-type Bβ2. Bβ2 is more heavily phosphorylated than Bβ1, and Ser20-22 substitution reduces ³²P incorporation into Bβ2.

N-terminal serines regulate the subcellular localization of Bβ2

To examine the functional consequence of Bβ2 phosphorylation, we mutated serines 20-22 and threonine 25 to alanines to mimic the unphosphorylated state of these amino acids. We then expressed Bβ2-GFP fusion proteins in Hela cells, fixed cells for immunofluorescence labeling of mitochondria, and examined colocalization of Bβ2-GFP with mitochondria by measuring Pearson’s coefficients (PC = 1 is perfect colocalization). Wild-type Bβ2 colocalized well with mitochondria (PC= 0.46), whereas neutralization of a positive charge (R6A) [17] reduced PCs to levels similar to cytosolic Bβ1 (PC=0.14, Fig. 2A,B). Analysis of single Thr→Ala and Ser→Ala substitutions in the Bβ2 N-terminus revealed that only S21A affected the localization of Bβ2, resulting in a small but significant increase in targeting to mitochondria (PC=0.5, Fig. 2B). In contrast, alanine substitution of all three vicinal serines (SSS20AAA) resulted in a robust increase of Bβ2 recruitment to mitochondria (PC=0.64, Fig. 2A,B). Alanine substitution of Ser21 and 22 (SS21AA) was nearly as effective (Fig. 4A, and data not shown). To provide complementary evidence for phosphorylation regulating Bβ2’s subcellular localization, Ser20-22 were replaced with aspartic acid to mimic phospho-serine. Phospho-mimetic substitution of two (SS21DD) or three serines (SSS20DDD) rendered Bβ2 completely cytosolic (PC ≈ Bβ1, Fig. 2A,B). Since Bβ2 SSS20DDD sometimes formed non-mitochondrial aggregates in cells, we instead analyzed Bβ2 SS21DD in subsequent experiments.

Fig. 2.

N-terminal serines influence subcellular localization of Bβ2. (A,B) The indicated Bβ-GFP proteins (green) were expressed in HeLa cells, and colocalization with mitochondria (cytochrome oxidase II antibody, red) was determined by epifluorescence microcopy. Representative images (A) show that wild-type (WT) Bβ2 has a mixed cytosolic/mitochondrial distribution, whereas dephospho (SSS20AAA) and phospho (SS21DD) Bβ2 are largely mitochondrial and cytosolic, respectively. Colocalization with mitochondria is quantified in (B) as the Pearson’s coefficient (mean ± SEM of ca. 400 cells from typically three independent experiments). (C) HEK293 cells expressing the indicated GFP-tagged B subunits were fractionated into membrane and cytosolic proteins and immunoblotted for the indicated antigens. Percent mitochondrial localization (% mito.) of B subunits was calculated as the ratio of mitochondrial to total (mitochondrial plus cytosolic) signals normalized to input signals. Statistics: unpaired Student’s t-test compared to Bβ2 wild-type; **p < 0.01, ***p < 0.001.

Fig. 4.

N-terminal phosphorylation of PP2A/Bβ2 influences mitochondrial morphology and Drp1 S656 dephosphorylation. (A,B) Hela cells were transfected with Bβ2-GFP (WT, SS21AA, and SS21DD) and OMM-PKA, fixed, and immunofluorescently labeled for cytochrome oxidase II (red). Mitochondrial morphology was determined from epifluorescence micrographs (representatives in (A)) and is expressed as form factor (circular mitochondria = 1) ((B), mean ± SEM of 57, 24, and 29 cells). (C-F) Drp1 S656 phosphorylation levels were assessed by phospho-specific antibody in COS1 cell lysates expressing GFP-Drp1 and the indicated Bβ subunits. Phospho S656 Drp1 signals were boosted by PKA activation via forskolin (2-7.5μM) and rolipram (2 μM) treatment for 60 min prior to cell lysis. Cells in (E,F) additionally received vehicle or FK506 (2 μM, 60 min) to inhibit calcineurin. (C,E) show representative immunoblots and (D,F) show quantification as the ratio of phospho- to total Drp1 (mean ± SEM of 6 (D) and 5 (F) experiments). Statistics: (B,D) unpaired Student’s t-test compared to wild-type Bβ2; (F) one-way analysis of variance (ANOVA) followed by pairwise tests with Bonferroni adjustments; *p < 0.05, **p < 0.01, ***p < 0.001.

To support the microscopy studies by biochemistry, we fractionated transfected HEK293 cells into cytosol and heavy membranes (including mitochondria), assessing fraction purity by immunoblotting for β-tubulin and TOM40. Twice as much Bβ2 than Bβ1-GFP or B’β-GFP was found in the heavy membrane fraction. 20% mitochondrial association is likely a low estimate for Bβ2, since the protein interacts weakly and transiently with receptor components of the TOM complex and dissociates from mitochondria during the fractionation process[16, 17]. Dephospho-Bβ2 (SSS20AAA) displayed enhanced association with the membrane fraction, while pseudophospho-Bβ2 (SS21DD) had a fractionation profile similar to cytosolic Bβ1 and B’β (Fig. 2C). Together, these observations suggest that Ser phosphorylation counteracts positive charges in the N-terminal mitochondrial import sequence to maintain cytosolic localization of Bβ2.

Autodephosphorylation of the PP2A/Bβ2 N-terminus is necessary for OMM translocation

We reasoned that intrinsic PP2A/Bβ2 phosphatase activity may mediate dephosphorylation of the Bβ2 N-terminus. Guided by the crystal structure of PP2A/Bα, Bα residues E27 and D197 were identified as critical for tau dephosphorylation by PP2A [19]. Because these residues are conserved in Bβ2, we hypothesized that they might regulate activity-dependent subcellular distribution of the PP2A/Bβ2 holoenzyme. To address this hypothesis, we mutated these amino acids (Bβ2 E26R and D196K) as well as a third residue that forms the tip of a loop extending from Bβ2 to contact the catalytic site on the C subunit (K87A, Fig. 3A). If PP2A/Bβ2 mediated its own N-terminal dephosphorylation, then mutation of residues involved with the phosphatase reaction should result in reduced mitochondrial localization. Indeed, substitution of E26 and K87 resulted in a dramatic loss of mitochondrial localization of Bβ2; however, the D196K substitution had no effect (Fig. 3B,C).

Fig. 3.

Mitochondrial localization of PP2A/Bβ2 requires intrinsic phosphatase activity. (A) Space filling model of PP2A/Bα (PDB 3DW8) highlighting residues that align with Bβ2 residues examined in this study. (B,C) Colocalization of Bβ2-GFP with mitochondria in HeLa cells was assessed as in Fig. 2. The SSS20AAA mutation rescues mitochondrial targeting of catalytically impaired (K87A) Bβ2. (D) Immunoprecipitation shows that Bβ2 mutations do not affect association with the PP2A catalytic subunit. Statistics: unpaired Student’s t-test compared to wild-type Bβ2; *p < 0.05, ***p < 0.001.

If substrate binding residues mediated mitochondrial localization of Bβ2 by allowing for dephosphorylation of the N-terminus, then dephospho Bβ2 (SSS20AAA) should be refractory to mutation of these residues. As expected, SSS20AAA, K87A double-mutant Bβ2 displayed constitutive mitochondrial association indistinguishable from Bβ2 SSS20AAA (Fig. 3B,C). Coimmunoprecipitation with the PP2A catalytic subunit confirmed that the Bβ2 mutations did not disrupt PP2A heterotrimer formation (Fig. 3D). These results suggest that autodephosphorylation of Ser20,21,22 is necessary for OMM translocation of Bβ2. Autodephosphorylation could occur either within the same PP2A/Bβ2 heterotrimer (intra-complex) or between holoenzymes (inter-complex).

N-terminal phosphorylation of Bβ2 impacts Drp1 dephosphorylation and mitochondrial shape

Mitochondrial fragmentation by Bβ2 overexpression requires recruitment of PP2A holoenzymes to mitochondria [11, 12]. To determine the effects of the non-phosphorylatable and pseudo-phosphorylated Bβ2 mutants on mitochondrial morphology, we performed digital morphometric analyses of transfected and fixed Hela cells. Compared to untransfected cells or the cytosolic Bβ1 subunit, expression of wild-type Bβ2 resulted in a reduction in the form factor of mitochondria indicative of enhanced mitochondrial fragmentation (Fig. 4A,B). Blocking Bβ2 phosphorylation with the SS21AA substitution amplified this effect, while pseudo-phosphorylation of Bβ2 (SS21DD) increased form factors to control levels, consistent with impaired mitochondrial fragmentation (Fig. 4A,B).

We recently reported that reversible phosphorylation of Drp1 at S656 regulates dendritic outgrowth and synapse formation in cultured hippocampal neurons [12]. Epistasis experiments with Ser656-mutant Drp1 indicated that both mitochondrial fusion by the OMM-targeted PKA/AKAP1 complex and mitochondrial fission by PP2A/Bβ2 require Drp1 S656 [9, 12]. To investigate whether PP2A/Bβ2 can dephosphorylate S656 of Drp1 in cells, we coexpressed in COS1 cells PP2A/B subunits along with a construct that replaces the endogenous Drp1 protein with a GFP-tagged version [8], allowing us to assess Drp1 phosphorylation in the transfected cell population only. To increase signals and thus allow for more robust quantification with the phospho-S656 antibody, Drp1 phosphorylation was boosted by treating cells with the adenylate cyclase agonist forskolin and the phosphodiesterase inhibitor rolipram 60 min prior to lysis. Compared to empty vector, expression of Bβ1 had no effect on Drp1 phosphorylation; however, wild-type Bβ2 reduced phospho-S656 Drp1 immunoreactivity by 34%. Rendering Bβ2 non-phosphorylatable (SSS20AAA) and pseudo-phosphorylated (SS21DD) augmented and abrogated, respectively, Drp1 dephosphorylation by Bβ2. Mutation of the Lys residue that contacts the catalytic site (K87A) also attenuated Bβ2 activity towards Drp1 S656 (Fig. 4C, D).

CaN was previously shown to dephosphorylate Drp1 at Ser656 [8, 10]. Furthermore, mitochondrial dysfunction and fragmentation in cells deficient in PINK1 (PTEN-induced kinase 1), a mitochondrial protein kinase mutated in familial Parkinson’s disease, was proposed to enhance CaN-mediated Drp1 dephosphorylation via increased [Ca²⁺]_i [20]. To investigate whether Drp1 S656 dephosphorylation involves a PP2A/Bβ2 → CaN “phosphatase cascade”, we inhibited CaN pharmacologically in transfected COS1 cells. Treatment with the CaN inhibitor FK506/tacrolimus resulted in an overall increase in Drp1 S656 phosphorylation, but ectopic expression of Bβ2 still dephosphorylated Drp1 compared to Bβ1 (Fig. 4E,F). These data indicate that CaN and mitochondria-targeted PP2A/Bβ2 dephosphorylate Drp1 independently.

We performed in vitro phosphatase assays to confirm that PP2A/Bβ2 can dephosphorylate Drp1 directly. To this end, PP2A complexes were immunoisolated from COS1 cells transfected with B regulatory subunits carrying C-terminal FLAG-GFP epitopes [16]. Using a model phosphopeptide substrate (RRA(pT)VA), rates of phosphate release were equivalent for Bβ1 and Bβ2-containing PP2A heterotrimers, but 2 times greater for PP2A/B’β after normalization for catalytic subunit levels (Fig. 5A). To assay Drp1 dephosphorylation, GST-Drp1_582-736 was phosphorylated with PKA and [γ-³²P]ATP at S656 [8]. PP2A/Bβ1 and Bβ2 dephosphorylated phospho-S656 Drp1 equally well, while PP2A/B’β activity was 4-fold lower (Fig. 5B). These results demonstrate that Drp1 S656 is a preferred substrate for PP2A heterotrimers containing Bβ (and perhaps other B-family) regulatory subunits. Because PP2A/Bβ2 is a better Drp1 phosphatase than PP2A/Bβ1 in intact cells, but not in vitro, we further conclude that the divergent N-terminus of Bβ2 mediates specific dephosphorylation of Drp1 via OMM localization of PP2A/Bβ2, rather than via direct substrate recognition.

Fig. 5.

PP2A/Bβ2 dephosphorylates Drp1 in vitro. (A,B) PP2A holoenzymes containing the indicated FLAG-GFP tagged regulatory subunits were immunoisolated from transfected COS1 and assayed for dephosphorylation of a model phosphopeptide (A) or GST-Drp1_582-736 that had been in vitro ³²P-phosphorylated on S656 by PKA (B, 15 and 45 min assay time). Raw phosphatase activities were adjusted for relative PP2A catalytic subunit levels obtained by immunoblotting (inset in (A)). Shown are means ± SEM of 5 experiments ((A), normalized to Bβ2) and means ± SEM of quadruplicate reactions from a representative experiment (B). Statistics: unpaired Student’s t-test compared to Bβ2; **p < 0.01.

Bβ2-induced neuronal death is modulated by N-terminal phosphorylation

We next assessed the ability of N-terminal phosphorylation to modulate mitochondrial localization of Bβ2-GFP in dendrites of primary hippocampal neurons 24 h after transfection. Compared to the mixed cytosolic/mitochondrial distribution of wild-type Bβ2, Bβ2 SS21DD was largely excluded from areas of high mitochondrial content (Fig. 6A). In contrast, Bβ2 SSS20AAA colocalized with dendritic mitochondria; most strikingly so in dying neurons with fragmented mitochondria (Fig. 6A, bottom right).

Fig. 6.

N-terminal phosphorylation modulates PP2A/Bβ2 subcellular localization and survival in neurons. (A) Primary hippocampal neurons were transfected as indicated and GFP-positive neurons were imaged after labeling mitochondria with an antibody to TOM20 (mito, red). (B,C) Hippocampal neurons were scored for viability 72 h after transfection by examining nuclear morphology, neurite integrity, and propidium iodide (PI) exclusion. (B), representative images; (C) quantification of neuronal death normalized to wild-type Bβ2-induced death (~40%) as means ± SEM from 3-4 experiments. (D) Model (see discussion). Statistics: (B) one-way analysis of variance (ANOVA) followed by pairwise test with Bonferroni adjustments; *p < 0.05, **p < 0.01, ***p < 0.001.

We previously showed that high-level overexpression of Bβ2 in cultured hippocampal neurons can induce apoptosis outright [11]. We therefore investigated whether phosphorylation of N-terminal serines can modulate neuronal survival by scoring Bβ2-GFP-positive neurons for integrity of the nucleus and neurites. At 72 hours post transfection, expression of wild-type Bβ2 increased basal cell death by two-fold (to ~40%), whereas pseudo-phosphorylated Bβ2 (SS21DD) was no more lethal than Bβ1. Conversely, dephospho Bβ2 (SSS20AAA) killed neurons more effectively than wild-type Bβ2 (Fig. 6B,C).

Discussion

The goal of this study was to understand how translocation of PP2A/Bβ2 from the cytosol to mitochondria is regulated and whether Bβ2 translocation modulates Drp1 activity and neuronal survival. Metabolic labeling identified serines 20-22 in the N-terminal mitochondrial targeting sequence of Bβ2 as phosphorylation sites in intact cells. Phosphomimetic substitutions of these residues 1) maintain Bβ2 in the cytosol, 2) inhibit Drp1 dephosphorylation at S656, 3) attenuate mitochondrial fragmentation, and 4) abolish the pro-apoptotic activity of PP2A/Bβ2 in hippocampal neurons. Since expression of constitutively dephosphorylated Bβ2 (SSS20AAA) resulted in opposite, gain-of-function phenotypes, we conclude that reversible phosphorylation of Bβ2 at Ser20-22 is an important regulatory mechanism in Drp1-mediated mitochondrial fission in neurons.

According to the model shown in Fig. 6D, PP2A/Bβ2 holoenzymes in healthy, unstressed neurons are sequestered in the cytosol via phosphorylation by as yet unidentified, pro-survival kinases. Upon cell stress, such as excitotoxic glutamate treatment or bioenergetic impairment with rotenone [11], Ser20-22 phosphorylation levels drop to reveal the net positive charge of Bβ2’s mitochondrial signal sequence, allowing for accumulation of the PP2A holoenzyme near the TOM import complex, in turn driving Drp1 activation by dephosphorylation, mitochondrial fission, and ultimately cell death. Several examples of phosphorylation enhancing mitochondrial protein import have been reported [21-24]. To the best of our knowledge, the present study is the first to demonstrate inhibition of mitochondrial targeting by phosphorylation of a leader sequence. Whether stress-mediated dephosphorylation of the Bβ2 N-terminus is a consequence of kinase inhibition or enhanced inter- or intramolecular autodephosphorylation of the PP2A/Bβ2 holoenzyme is an important question that deserves further study.

We also provide evidence that a pivotal inhibitory phosphorylation site in Drp1, S656 [7, 8], is targeted not only by CaN, but also by PP2A/Bβ2. PP2A/Bβ2-mediated Drp1 dephosphorylation is independent of CaN in intact cells, and both Bβ splice variants, but not a structurally unrelated PP2A regulatory subunit efficiently dephosphorylate Drp1 S656 in vitro. Intriguingly, PP2A inhibition by okadaic acid or calyculin A does not increase Drp1 S656 phosphorylation levels in PC12 cells, which express most PP2A regulatory subunits, but not Bβ2 [8]. Since point mutations that block mitochondrial targeting of Bβ2 also prevent Drp1 dephosphorylation, we conclude that PP2A-mediated dephosphorylation and activation of Drp1 strictly depends on mitochondrial localization of PP2A by Bβ2. Because cytosolic (Bβ1) and mitochondrial (Bβ2) splice forms mediate equally efficient Drp1 dephosphorylation in vitro, local concentration at the OMM most likely accounts for specific dephosphorylation of Drp1 by the PP2A/Bβ2 holoenzyme in vivo.

While PP2A/Bβ2 increases susceptibility to neuronal insults as shown by knockdown and overexpression approaches [11], the phosphatase also plays important physiological roles, stimulating mitochondrial division to foster synaptogenesis and curb dendritic hyperplasia in cultured hippocampal neurons [12]. Regulation of OMM translocation of PP2A/Bβ2 by N-terminal phosphorylation therefore provides neurons with temporal and spatial control of mitochondrial morphogenesis in development, plasticity, and survival.

Materials and methods

Reagents and cDNA constructs

The following commercial antibodies were used: mouse anti-GFP clone 86/8 (NeuroMab), rabbit anti-GFP (ab290, Abcam), mouse anti-MTCO2 (cytochrome oxidase subunit II, Neomarkers), mouse anti-Drp1 (BD Transduction Laboratories), mouse anti β-tubulin (E7, Developmental Studies Hybridoma Bank), rabbit anti-TOM40 (Santa Cruz Biotechnology). Mouse anti phospho-S656 Drp1 was raised at the Iowa State Hybridoma Facility [9]. For immunofluorescence staining, we purchased Alexa fluorophore-coupled secondary antibodies (Invitrogen). For quantitative immunoblot analysis, infrared fluorophore-coupled secondary antibodies were purchased from LI-COR Biosciences (Lincoln, NE). The GFP-Drp1 [8], Bβ1-, Bβ2-, B’β-FLAG-GFP [16, 17, 25] and OMM-PKA [9] vectors were previously described. Bβ2 point mutants were generated by PCR-based methods taking advantage of a nearby, unique EcoRI site.

In vitro phosphorylation and metabolic labeling

For in vitro assays, COS1 cells were transiently transfected with Bβ2 N-terminus-FLAG-GFP fusion proteins using Lipofectamine 2000 [17]. After 48 h, proteins were immunoisolated with FLAG-directed antibodies [16] and in vitro phosphorylated (30 min, 30°C) with 0.5 μM CaMKIIα, 2 μM calmodulin (kind gift of Roger Colbran, Vanderbilt Univ.) in buffer containing 25 mM HEPES (pH 7.4), 2 mM CaCl₂, 10 mM MgCl₂, 0.2 μCi/μl [γ-³³P]ATP, and 200 μM unlabeled ATP. Kinase reactions were stopped with 25 mM EDTA and analyzed by SDS-PAGE, total protein stain, and phosphorimager.

For metabolic ³²P labeling, COS1 cells were transfected with Bβ splice variants carrying C-terminal FLAG or V5 epitopes or Bβ2_1-35-FLAG-GFP. 24 hours later the cells were washed once with phosphate-free RPMI (MP Biomedical) and incubated in media containing phosphate-free RPMI, 1% dialyzed FBS, 0.5 mCi/ml ³²P-labeled ortho-phosphate (PerkinElmer) for 3.5 hours at 37°C. When full-length Bβ was expressed, 25 nM calyculin A was added to the media for the last 30 min. Cells were washed with phosphate-free RPMI and Bβ2_1-35-GFP was immunoprecipitated with anti-FLAG or anti-V5 antibodies in the presence of phosphatase inhibitors as described [8]. Immunoprecipitates were analyzed by phosphorimager and by immunoblotting for GFP, FLAG, or V5 tags.

Subcellular fractionation and Drp1 S656 phosphorylation analysis

For subcellular fractionation, HEK293 cells expressing GFP-tagged B subunit were permeabilized in 0.5 mg/ml digitonin, 20 mM HEPES pH 7.4, 1 mM EDTA, 1 mM EGTA, 1 mM benzamidine, 5 mg/ml leupeptin, 1 mM dithiotreitol (DTT), and 1 mM phenylmethylsulfonyl fluoride (PMSF). Lysates were centrifuged (10 min, 20,000 × g, 4°C) to separate the cytosolic and heavy membrane fractions essentially as described [9].

To quantify Drp1 phosphorylation in cells, COS cells were cotransfected with GFP-tagged B subunits or empty vector and GFP-Drp1 at 1:1 plasmid mass ratio using Lipofectamine 2000. The GFP-Drp1 plasmid replaces endogenous Drp1 with the GFP-tagged protein by co-expression of RNAi-resistant cDNA and H1 promoter-driven shRNA [8]. After 24 h, cells were treated with forskolin/rolipram (10/1 μM, 1 h) to upregulate PKA activity, lysed in SDS sample buffer containing 2 mM EDTA and 1 mM microcystin, and sonicated with a probe tip to shear DNA.

Subcellular fractions or total cell lysates were resolved on 10% polyacrylamide gels and transferred to nitrocellulose membrane. Blots were probed with primary and infrared fluorophore-labeled secondary antibodies, and bands were visualized using a LI-COR Odyssey infrared fluorescence scanner. Band intensities were quantified using the ImageJ gel analysis macro set. Phosphorylation of GFP-Drp1 was quantified as the ratio of phospho-Ser656 Drp1 antibody [8] to GFP antibody immunoreactivity of the same band.

Mitochondrial colocalization and morphology analysis

Hela cells were cultured on No. 1 cover glasses (20 mm² chamber, Nalgen Nunc) and transfected using LipofectAmine 2000. Cells were fixed with 4% parformaldehyde and subjected to immunofluorescence staining with antibodies to cytochrome oxidase II and GFP as reported [9]. Images were captured at 630X magnification using a Leica epifluorescence microscope and processed for local contrast enhancement (contrast-limited adaptive histogram equalization, CLAHE) using ImageJ. Images subjected to red/green colocalization analysis underwent a second processing step involving iterative 2D deconvolution (macro code and algorithm parameters are available upon request). Colocalization of GFP-tagged Bβ subunits with mitochondria was quantified as the Pearson’s coefficient using the JaCoP plug-in for ImageJ. To quantify mitochondrial morphology, mitochondria channel images were analyzed using a custom ImageJ macro described previously [9, 26].

GST-Drp1 purification, in vitro phosphorylation, and phosphatase assays

A plasmid expressing GST-Drp1_582-736 was created by ligation of a human Drp1 (NCBI accession# NP_036192) PCR fragment into pGEX-6P1 digested with BamHI and XhoI. GST-Drp1 was expressed in BL21/DE3 strain E. coli and purified on glutathione-agarose according to standard protocols. GST-Drp1_582-736 (13.5 μM) was specifically phosphorylated on S656 (30 min, 30°C) with 0.27 μM PKA catalytic subunit (kind gift of Susan Taylor, UC San Diego) in buffer containing 20 mM Tris (pH 7.5), 10 mM MgCl, 0.1 μCi/μl [γ-³²P]ATP, 100 μM unlabeled ATP, and 1 mM benzamidine. Reactions were arrested by the addition of 25 mM EDTA and free ATP was removed by two passages through desalting columns (Zeba, Thermo Sci.).

PP2A holoenzymes containing transfected and epitope-tagged regulatory subunits complexed with endogenous A and C subunits were immunoisolated from COS1 cells as described [16]. Phosphatase complexes on agarose beads were resuspended in reaction buffer containing 50 mM Tris (pH 7.5), 0.1% Triton X-100, 2 mM EDTA, 2 mM EGTA, 1 mM benzamidine, 0.05% β-mercaptoethanol and 2 mg/ml BSA. Phosphatase reactions were started by the addition of ³²P-labeled GST-Drp1_582-736 and incubated at 30°C with intermittent agitation. Reactions were stopped at 15 and 45 min by addition to 20% trichloroacetic acid. After centrifugation at 22,000 × g, acid-soluble ³²P was quantified by liquid scintillation counting. For peptide dephosphorylation experiments, phosphatase activities towards a phosphopeptide derived from myosin light chain (RRA(pT)VA, 100 μM) were determined using a molybdate:malachite green-based colorimetric assay (Promega).

Mitochondrial localization and survival assays in hippocampal neurons

Hippocampal neurons from E18 rat embryos were cultured in Neurobasal medium with B27 supplement (Invitrogen), transfected with GFP-tagged B regulator subunits using LipofectAmine 2000 (0.1%, 2 ug/ml DNA) at 10–14 days-in-vitro (DIV), and fixed with 3.7% paraformaldehyde 24 or 72 h later. For the localization experiments, cultures were immunofluorescently labeled for GFP to enhance intrinsic fluorescence of the Bβ-GFP proteins, and for the mitochondrial protein TOM20. Neurons were imaged at 1000X magnification using a Leica epifluorescence microscope. For survival assays, hippocampal cultures were fixed 72 hours after transfection and labeled with GFP-directed antibodies and Alexa 488-coupled secondary antibodies. Nuclei were stained with 1 μg/ml Hoechst 33342. In some experiments, cultures were also incubated with propidium iodide (1 μg/ml) 5 min prior to fixation in order to stain the nuclei of necrotic cells that lost membrane integrity. Death was quantified as the percentage of transfected neurons with condensed, irregular, or fragmented nuclei or dystrophic neurites by experimenters blinded to transfection conditions.

Statistical analysis

The unpaired Student’s t-tests were performed using Excel. The one-way analysis of variance (ANOVA) and the pairwise tests with Bonferroni adjustments were carried out with the LawStat library of the statistical software package R [27]. Unless indicated otherwise, data are representative of three or more independent experiments.

Abbreviations

AKAP1

A kinase anchoring protein 1

ANOVA

analysis of variance

CaN

calcineurin

Drp1

dynamin-related protein 1

OMM

outer mitochondrial membrane

PC

Pearson’s coefficient

PINK1

PTEN-induced kinase 1

PKA

protein kinase A

PP2A

protein phosphatase 2A

SCA12

spinocerebeller ataxia type 12

SEM

standard error of the mean

TOM

translocase of outer membrane