Microtubule Affinity-regulating Kinase 2 (MARK2) Turns on Phosphatase and Tensin Homolog (PTEN)-induced Kinase 1 (PINK1) at Thr-313, a Mutation Site in Parkinson Disease

Background: Deregulation of protein kinases MARK2 and PINK1 has been linked to neurodegenerative diseases. Results: MARK2 was identified as an upstream regulator of PINK1. Conclusion: The MARK2-PINK1 cascade provides new insights into the regulation of mitochondrial trafficking in neurons. Significance: This protein kinase signaling axis might provide a link between neurodegenerative processes in Alzheimer and Parkinson diseases. The kinase MARK2/Par-1 plays key roles in several cell processes, including neurodegeneration such as Alzheimer disease by phosphorylating tau and detaching it from microtubules. In search of interaction partners of MARK2, we identified phosphatase and tensin homolog (PTEN)-induced kinase 1 (PINK1), which is important for the survival of neurons and whose mutations are linked to familial Parkinson disease (PD). MARK2 phosphorylated and activated the cleaved form of PINK1 (ΔN-PINK1; amino acids 156–581). Thr-313 was the primary phosphorylation site, a residue mutated to a non-phosphorylatable form (T313M) in a frequent variant of PD. Mutation of Thr-313 to Met or Glu in PINK1 showed toxic effects with abnormal mitochondrial distribution in neurons. MARK2 and PINK1 were found to colocalize with mitochondria and regulate their transport. ΔN-PINK1 promoted anterograde transport and increased the fraction of stationary mitochondria, whereas full-length PINK1 promoted retrograde transport. In both cases, MARK2 enhanced the effects. The results identify MARK2 as an upstream regulator of PINK1 and ΔN-PINK1 and provide insights into the regulation of mitochondrial trafficking in neurons and neurodegeneration in PD.

Transport of vesicles and organelles is vital for cells, especially in neurons. Tau and other microtubule-associated proteins promote the assembly and stabilization of neuronal microtubule tracks and ensure microtubule-dependent transport. We showed previously that elevated tau leads to inhibition of traffic because of competition with motor proteins for micro-tubule binding. This inhibition can be relieved by MARK2, 4 which phosphorylates microtubule-associated proteins and causes their detachment from microtubules (1). Consequently, MARK2 has a pivotal role in the regulation of microtubule dynamics and cellular transport.
Kinases of the MARK/Par-1 family perform diverse functions in neuronal polarity, transport, migration, and neurodegeneration (2). MARK/Par-1 comprises several domains, which offer multiple means of regulation. For example, the catalytic domain of MARK2 is activated by phosphorylation of Thr-208 by MARKK/TAO1 or LKB1 and inactivated by phosphorylation of Ser-212 by GSK3␤ (3)(4)(5). MARK2 activity can be inhibited by complex formation (e.g. with PAK5 (6)) or intramolecular interactions between domains (7,8). Thus, in view of the importance of MARK for neuronal polarity and neurodegeneration, we set out to identify novel partners of MARK2 and herein describe the interaction of MARK2 with ⌬N-PINK1 (aa 156 -581)/full-length PINK1 (PINK1 FL ) and the effects of these kinases on mitochondrial transport and distribution.
PINK1 is a mitochondrially targeted serine/threonine kinase that promotes cell survival, particularly under conditions of oxidative/metabolic stress (9 -11). PINK1 mutations are linked to PD and mostly result in loss of kinase activity. The molecular events responsible for neuronal death as well as physiological substrates or regulators of PINK1 are currently a matter of debate (12,13). The mitochondrial localization of PINK1 (14,15) and alterations in mitochondrial morphology, dynamics, and function caused by PINK1 deficiency (16 -20) implicate a central role of PINK1 in normal mitochondrial function. But the location of PINK1 is still unclear. PINK1 has been reported to reside either in the inner mitochondrial membrane, the mitochondrial intermembrane space, the outer mitochondrial * This work was supported in part by the Deutsche Forschungsgemeinschaft, membrane (21,22), or even the cytoplasm (15,23). Other data support a model in which PINK1 spans the outer mitochondrial membrane with the N-terminal end inside the mitochondria and the C-terminal kinase domain facing the cytosol (24). This topology relies on a transmembrane domain located just after the mitochondrial targeting sequence (MTS) of PINK1 and allows substrate phosphorylation outside of the mitochondria. However, PINK1 protein is cleaved upon entry into the mitochondria to produce two N-terminally truncated protein fragments of 54 and 45 kDa (12) that localize more to the cytosolic than to the mitochondrial fraction (25). This suggests that cytosolic targets and signal transduction pathways may be modified by cleaved PINK1 to affect neuronal survival (23). The physiological function of these truncated PINK1 proteins is still unknown, although ⌬N-PINK1 is quite stable, indicating that this is one of the predominant forms in the steady state (25). Here, we show that MARK2 interacted with and phosphorylated preferentially the 45-kDa fragment (termed ⌬N-PINK1), thereby increasing its kinase activity. Both kinases colocalized with mitochondria in primary neurons, especially in the growth cone, and modified mitochondrial transport. The main site phosphorylated by MARK2 on ⌬N-PINK1 was Thr-313, a frequent mutation site (T313M) in certain forms of familial PD (26). This suggests that failure of the MARK2-PINK1 signaling cascade contributes to the disease.

EXPERIMENTAL PROCEDURES
Plasmids-The full-length cDNA clone PINK1 wild-type (WT) pCMV-Sport6 was obtained from the RZPD (Ressourcenzentrum Primärdatenbank, Berlin, Germany). PINK1 FL was amplified by PCR using oligonucleotides that introduce an NdeI restriction site at the start codon and an NheI restriction site behind the stop codon. The 45-kDa fragment ⌬N-PINK1 was chosen to include residues 156 -581 because this corresponds closely to the cellular fragment (whose exact cleavage site is not known (27)). The fragment was cloned into pEU-myc and examined by sequencing. PINK1 deletion and point mutants (e.g. ⌬N-PINK1, aa 156 -581) were generated by PCR using primers containing appropriate mutations and restriction sites. All PINK1-YFP constructs were generated by PCR using oligonucleotides that introduce a HindIII restriction site at the start codon and an NheI restriction site behind the stop codon of the PINK1 sequence. The fragment was cloned into the YFP-pShuttle fluorescence expression vector to generate adenovirus as described previously (28). PINK1 FL -pVL-His was generated for recombinant expression in Sf9 cells using a modified pVL1393 vector (BD Pharmingen (3)) and the NdeI/NheI PINK1 FL fragment. ⌬N-PINK1-pGEX-4T-3 was generated via blunt-end cloning to express and purify recombinant protein.
Yeast Two-hybrid Analysis-The yeast two-hybrid screen and direct binding assays were performed according to the manufacturer's instructions for the MATCHMAKER GAL4 Two-Hybrid System (Clontech) and were described previously (6).
Immunoprecipitation and Immunoblot Analysis-Immunoprecipitation and immunoblot analysis were done by standard procedures as described previously (6).
Antibodies and Dyes-Antibodies, markers, immunofluorescence, and coupling methods are described in the supplemental Experimental Procedures.
Cell Fractionation-Preparations of mitochondrial fractions from N2A cells were kindly provided by A. M. Cuervo (Albert Einstein College, Bronx, NY) and are described elsewhere (29).
In Vitro Kinase Activity Assay-Purified PINK1 and MARK2 proteins were used at a concentration of 1 M each. 150 M histone H4 (Sigma-Aldrich) was used as a substrate for ⌬N-PINK1. As substrate for MARK2, we used 150 M TR1 peptide derived from the first repeat of tau protein containing the KXGS motif (30). The activity was assayed as described previously (3). Phosphorylated TR1 peptide was measured in a scintillation counter (Tricarb 1900 CA, Packard), and phosphorylated proteins were separated by SDS-PAGE and visualized by autoradiography using a bioimaging analyzer (BAS 3000, Raytest, Straubenhardt, Germany).
Preparation of in Vitro Phosphorylated PINK1 Peptides-The preparation of the in vitro phosphorylated PINK1 peptides is described in the supplemental Experimental Procedures.
Mass Spectrometry-MALDI-TOF mass spectrometry was performed on a Voyager-DE STR time-of-flight instrument (Applied Biosystems, Darmstadt, Germany) equipped with a nitrogen laser operating at 337 nm. The peptides were analyzed in a positive ion delayed extraction reflector mode using ␣-cyano-4-hydroxycinnaminic acid (Sigma-Aldrich) (10 mg/ ml in 0.05% TFA with 50% acetonitrile) as the UV light-absorbing matrix. Samples were dissolved in matrix solution and spotted onto the target plate, allowing them to air dry. All mass spectra were externally calibrated with standard peptides (New England Biolabs, Frankfurt, Germany) containing angiotensin, neurotensin, and various ACTH fragments. All spectra were analyzed using Data Explore TM software (Applied Biosystems).
Cell Culture-Cell culture was performed with HEK293, CHO, PC12, and Sf9 cells following standard protocols at 37°C with 5% CO 2 . Details are described in the supplemental Experimental Procedures.
Immunostaining, Labeling, and Live Cell Light Microscopy-Cells were fixed in 3.7% paraformaldehyde for 15 min at room temperature. Cells were then washed twice with PBS to remove paraformaldehyde. Afterward, cells were permeabilized with 80% ice-cold methanol for 5 min at Ϫ20°C and washed three times with PBS. Cells were incubated with primary antibody for 1 h at 37°C after blocking with 10% goat serum for 30 min at 37°C to avoid unspecific binding. Cells were washed three times before incubating with fluorescence-labeled secondary antibody for 1 h at 37°C. Cells were then mounted with Permafluor (Beckman Coulter, Krefeld, Germany) and observed with a 63ϫ oil immersion objective on an LSM510 Meta confocal microscope (Zeiss, Jena, Germany) using lasers, beam splitters, and filters according to the fluorophores. For time lapse imaging of mitochondria, cells were plated on glass bottom dishes. Cells were kept at 37°C by air heating and supplied with 5% CO 2 . Images were taken every 2-15 s adapted to the best experimental settings for tracking of mitochondria using a 63ϫ objective, beam path, and laser settings for Cy5 for mitochondrial observation and YFP and CFP for confirming transfection of retinal ganglion cells (RGCs).
Data Analysis-The Zeiss LSM510 Image Browser software was used to track individual particles. We interpreted all stained mitochondria that were transported toward the axonal growth cone as anterograde and those that moved to the cell body as retrograde. Kymographs were then generated and analyzed using ImageJ Kymograph. The observation time varied from 5 to 10 min. To subtract the contribution from random Brownian motions, a mitochondrion was considered as actively transported if its velocity was faster than 0.3 m/s (32). The following parameters were measured: velocity v (m/s), distance of transport (run length L; m), direction (anterograde or retrograde), and pausing time. Mitochondria with velocities below 0.3 m/s were defined to be pausing. If a mitochondrion stood idle for more than 80s it was classified as stationary. Each movement (anterograde, retrograde, pause, or stationary) was counted as one independent value. For each transfection, three independent movies were evaluated. The averages were analyzed for statistical significance by one-or two-way analysis of variance. N indicates the number of experiments, and n is the number of observed mitochondrial movements. p values of Ͻ0.05, Ͻ0.01, and Ͻ0.001 are indicated with one, two, or three asterisks, respectively.

Identification of PINK1 as MARK2-interacting Protein-To
identify new regulators or alternative substrates of MARK2, a human fetal brain cDNA library was screened using the MATCHMAKER yeast two-hybrid system. The bait was obtained by cloning the complete coding sequence of a kinasedead MARK2 cDNA mutant into the two-hybrid vector. The MARK2 T208A/S212A mutation was chosen because it mimics the unphosphorylated state of the kinase and keeps the substrate cleft in an open conformation (3,8). Positive clones were fur-ther analyzed by DNA sequencing. One insert contained the DNA encoding part of the PINK1 kinase domain (aa 320 -449). This suggested a possible interplay between these kinases in neurodegeneration because mutations in PINK1 are related to early onset Parkinsonism (33), and MARK2 is implicated in Alzheimer disease (2). Fig. 1 illustrates the interactions between ⌬N-PINK1 and MARK2. Different domains and mutants of MARK2 were . Empty pEU vector (lane 1) was used to make the total amount of transfected DNA equivalent. Cell lysates were immunoprecipitated with anti-HA antibody (row 3) and immunoblotted either with anti-HA antibody (row 1, WB: MARK) or anti-myc antibody (rows 2 and 3, WB: PINK1). ⌬N-PINK1 immunoprecipitates more strongly with the inactive MARK2 mutant than with MARK2 WT probably due to altered conformation after activation. In the case of PINK1 FL , only a non-phosphorylatable form (T313M) binds sufficiently well to MARK2 T208A/S212A . IP, immunoprecipitation; WB, Western blotting. checked in combination with the detected PINK1 fragment in direct two-hybrid interaction tests to map the binding domains of the kinases (Fig. 1A). Both proteins interact via their kinase domains (Fig. 1B). The C-terminal domain of MARK2 is not involved in binding the PINK1 fragment (aa 320 -449) because this domain alone showed no interaction with the PINK1 construct. To confirm the results of the direct two-hybrid tests, a coimmunoprecipitation using HEK293 cells was carried out (Fig. 1C). We generated ⌬N-PINK1 (residue 156 -581) and different mutants to study its activity and interactions in the absence of possible influence by its adjacent domains (34). This construct corresponds closely to the most stable 45-kDa PINK1 cleavage product (25). We coexpressed wild-type or kinasedead HA-tagged MARK2 in combination with ⌬N-PINK1 or PINK1 FL . When lysates of these cells were immunoprecipitated with anti-HA antibody, ⌬N-PINK1 was detected in the HA-MARK2 immune complexes (Fig. 1C, row 3, lanes 5 and 6). The interaction with kinase-dead MARK2 was stronger than with MARK2 WT , indicating that conformational changes of MARK2 in its activation loop may enhance the binding to ⌬N-PINK1. An interaction of PINK1 FL and MARK2 was not detectable by coimmunoprecipitation (data not shown). However, coexpression of kinase-dead HA-tagged MARK2 and a non-phosphorylatable form of PINK1 FL (T313M mutant) led to binding of both kinases detectable by coimmunoprecipitation (Fig. 1C, row 3, lane 9), suggesting that the phosphorylation reaction modulates the strength of the interaction (for further details, see Figs. 3

and 4).
To examine the subcellular localization of MARK2 and ⌬N-PINK1, primary rat cortical neurons were transfected with ⌬N-PINK1-YFP and CFP-MARK2 T208A/S212A by adenovirus and analyzed by immunofluorescence (Fig. 2). Both kinases have a vesicular appearance in axons and dendrites and a mainly cytoplasmic distribution in the cell soma. A strong colocalization was visible particularly in vesicle-like structures, confirming the two-hybrid and coimmunoprecipitation data (Figs. 2 and 4 -6). Cortical neurons transfected with PINK1 FL and MARK2 T208A/S212A showed partial colocalization due to different expression patterns of PINK1 FL and ⌬N-PINK1 (Figs. 2 and 7-9). PINK1 FL was more localized in the cell soma and to a lesser extent in axons and dendrites.
MARK2 Phosphorylates ⌬N-PINK1 at Thr-313 and Enhances Its Activity-Because the catalytic domains of the two kinases interact with one another, the question arose whether this influences their kinase activities. To examine this possibility, we performed in vitro phosphorylation assays to test the ability of MARK2 to phosphorylate PINK1 FL or ⌬N-PINK1 and vice versa. First, bacterially purified PINK1 FL or ⌬N-PINK1, alone or in combination with MARK2 WT , and constitutively active MARK2 T208E (3) were incubated with [ 32 P]ATP. PINK1 FL or MARK2 WT alone showed only slight autophosphorylation (Fig. 3A, lanes 1 and 4) that became strong in the case of MARK2 T208E (Fig. 3A, lane 3). ⌬N-PINK1 revealed no autophosphorylation (Fig. 3A, lane 2), but in combination with MARK2 T208E , it was strongly phosphorylated (Fig. 3A, lane 6 and bar 6). The effect on PINK1 FL was weak (Fig. 3A, bar 5), indicating a strong preference of MARK2 for the cleaved (cytoplasmic) ⌬N-PINK1 or a conformational hindrance of PINK1 FL in vitro. However, the results of the coimmunoprecipitation experiment with MARK2 T208A/S212A and the non-phosphorylatable PINK1 FL mutant indicate that MARK2 binds to the mutant but cannot complete the phosphorylation reaction in this situation. Neuronal cultures were transfected with MARK2 T208A/S212A and ⌬N-PINK1 adenoviruses, cultured for 24 h, and fixed. We chose the adenoviral expression system because the protein expression is weak, and the properties of the endogenous kinases are reflected by the exogenous kinases. The intracellular distribution of both proteins was visualized by fluorescence microscopy. A strong colocalization of ⌬N-PINK1-YFP (yellow; panels 1 and 4) and CFP-MARK2 T208A/S212A (inactive) (cyan; panels 2 and 5) in vesicle-like structures can be observed in the axon and dendrites of the neuron (merge; panels 3 and 6, arrows). PINK1 FL -YFP (yellow; panels 7 and 9) is mainly expressed in the cell soma but also colocalizes with CFP-MARK2 T208A/S212A (cyan; panels 8 and 9) in axons and dendrites. Scale bar, 10 m.
To investigate the functional consequences of the phosphorylation of ⌬N-PINK1 by MARK2, we first identified the phosphorylation sites by phosphopeptide mapping followed by mass spectrometry. Constitutively active MARK2 T208E and ⌬N-PINK1 were incubated with [ 32 P]ATP. Phosphorylated ⌬N-PINK1 was separated by SDS-PAGE followed by tryptic in-gel digestion and phosphopeptide enrichment. These peptides were mapped by thin-layer electrophoresis (horizontal) followed by ascending chromatography (vertical). The main spot shown in Fig. 3B corresponds to phosphorylated Thr-313 of ⌬N-PINK1 peptides identified by MALDI-TOF MS analysis. The minor spots are in accordance with phosphorylated Thr-333 and Ser-335 of ⌬N-PINK1 peptides. Because T313M is known as a frequent PD-associated mutation in the kinase domain of PINK1 (26), this residue may have an important function in the regulation of PINK1. The Thr-313, which is phosphorylated by MARK2, is located in ␤-strand 5 of PINK1 (Fig. 3C, marked in green). Phosphorylation of this residue could result in the interaction of this strand with helix C or its insert, thus stabilizing helix C. This helix is part of the scaffold that fixes the Mg-ATP underneath the P-loop and is a requirement for the activity of the kinase.
Because mutation of threonine to alanine prevents further phosphorylation/activation, whereas mutation to phosphorylation-mimicking glutamate produces active forms of many kinases (5), several PINK1 mutants were created replacing Thr-313 to confirm the specificity of this newly identified phosphorylation site. Bacterially purified ⌬N-PINK1 T313A was incubated alone or in combination with MARK2 T208E in the presence of [ 32 P]ATP. Indeed, this mutation caused a reduction (ϳ60%) of the phosphorylation of ⌬N-PINK1 by active MARK2, but phosphorylation of ⌬N-PINK1 still occurred, indicating that MARK2 also phosphorylates other sites of the kinase (Fig. 4A, compare lanes/bars 7 and 9). Although Thr-313 was mutated to glutamate in ⌬N-PINK1, MARK2 T208E increased further the phosphorylation by 67% in comparison with the approach with non-mutated ⌬N-PINK1 (Fig. 4A, compare lanes/bars 7 and 8). This led to the assumption that Thr-313 is a "priming" phosphorylation site to prepare the ground for further phosphorylation of ⌬N-PINK1. was mapped by thin-layer electrophoresis (horizontal) followed by thin-layer chromatography (vertical). The main spot corresponds to peptides containing phosphorylated Thr-313, a residue known to be mutated to a non-phosphorylatable form in PD. The minor spots are in accordance with phosphorylated Thr-333 and Ser-335 of ⌬N-PINK1 peptides. The identified phosphorylated residues are indicated in the schematic representation of PINK1 FL . C, structural model of PINK1 by SwissModel using CaMK1 (Protein Data Bank code 1A06) as a template. The activation loop is shown in purple. Inserts that are not common in other kinases are gray; the catalytic aspartate is blue. Threonine 313, which is phosphorylated by MARK2, is located in ␤-strand 5 (green). Phosphorylation of this residue likely results in the interaction of strand 5 with helix C or its insert, thus stabilizing helix C. This helix is part of the scaffold that fixes the Mg-ATP underneath the P-loop and is a requirement for the activity of the kinase. CAT, catalytic domain; TM, transmembrane domain.
To clarify whether this strong phosphorylation of ⌬N-PINK1 by MARK2 T208E influences the activity of ⌬N-PINK1, phosphorylation assays using the artificial PINK1 substrate histone H4 were performed (34). The results indicate that MARK2 T208E alone phosphorylates histone H4 only slightly (Fig. 4B, bar 5), but in combination with ⌬N-PINK1, the substrate phosphorylation was enhanced (Fig. 4B, bar 7), indicating that MARK2 stimulates the kinase activity of ⌬N-PINK1. ⌬N-PINK1 T313E was not constitutively active (Fig. 4B, bar 3), but MARK2 T208E induced a 2.5-fold higher activation of the Glu mutant in comparison with the batch with non-mutated ⌬N-PINK1 (Fig. 4B,  bar 8). Accordingly, the ⌬N-PINK1 T313A mutant was not completely kinase-dead, but the activation could not be further enhanced by MARK2 T208E in comparison with the non-mutated ⌬N-PINK1 (Fig. 4B, bars 7 and 9). These results prove that phosphorylation of Thr-313 is important but not sufficient for full activation of ⌬N-PINK1.
To exclude the possibility that MARK2 is activated by ⌬N-PINK1 and thus causes the enhanced phosphorylation of histone, the kinase activity of MARK2 was measured by its phosphorylation of the tau peptide TR1 (derived from the first repeat sequence of tau protein (30)). Constitutively active MARK2 T208E and PINK1 FL or ⌬N-PINK1 were mixed together with the TR1 peptide and incubated with [ 32 P]ATP, and MARK2 activity was measured by the phosphorylation of TR1 using a scintillation counter. There was a ϳ30 -40% inhibition of MARK2 activity induced by PINK1 FL or ⌬N-PINK1 possibly due to binding of PINK1 FL and ⌬N-PINK1 (Fig. 4C, bars 4 and  5). A similar inhibitory mechanism was shown previously with MARK2 and PAK5 (6). PINK1 FL also reduced MARK2 activity, although the MARK2-induced phosphorylation of PINK1 FL was weak in vitro. To clarify whether PINK1 FL is linked to MARK2 directly or indirectly, a protein stability assay was performed. CHO cells were individually transfected with PINK1 FL alone or in combination with MARK2 T208A/S212A (Fig. 4D). At 16 h after the transfection, the cells were treated for 1 h with 10 M epoxomycin (proteasome inhibitor) or 50 M cycloheximide (inhibitor of protein biosynthesis). Both the proteasomal inhibitor and MARK2 T208A/S212A increased the stability of fulllength and cleaved PINK1, arguing for a physiological relevance of this kinase-substrate interaction. MARK2 T208A/S212A stabilized particularly PINK1 FL and the shortest cleaved PINK1 corresponding to our ⌬N-PINK1 (Fig. 4D, row 1, lane 4). Additionally, a phosphorylation assay was carried out. CHO cells were transfected with ⌬N-PINK1 or PINK1 FL alone or in combination with MARK2 T208E . PINK1 was immunoprecipitated, and phosphorylation was detected by immunoblotting with antiphosphoserine or anti-phosphothreonine antibody. As expected, ⌬N-PINK1 was mainly phosphorylated at threonine sites in a MARK2-dependent manner (supplemental Fig. S1, lane 5), whereas PINK1 FL coexpressed with MARK2 showed phosphorylation at serine-and threonine-specific sites (supplemental Fig. S1, lane 6). These results confirmed our assumption that even PINK1 FL is modified by MARK2. In summary, we identified MARK2 as an upstream activating kinase of PINK1.
MARK2 Colocalizes with ⌬N-PINK1 and Mitochondria in Primary Cortical Neurons-Endogenous PINK1 occurs both in the cytoplasm and on mitochondria despite the mitochondrial localization signal in PINK1 (23). But ⌬N-PINK1 can also be found at mitochondria (35,36). Nevertheless, the exact localization is still a matter of debate. MARK2 is a cytoplasmic kinase, but we had shown earlier that MARK2 has a dotted distribution with a diffuse background, suggesting that it is partially associated with vesicles (6, 37) (Fig. 2). To study the localization of both kinases and to determine in which cell compartment the interaction of both kinases occurs, we initially stained endogenous active MARK2 in primary cortical neurons using a phosphospecific MARK2 antibody (MARK Thr(P)-208) and costained for mitochondria (Fig. 5A, panels 1-6). Surprisingly, we found an extensive colocalization (up to 80%) of MARK2 not only with vesicles but also with mitochondria, particularly in axons and dendrites, emphasized at the higher magnification (Fig. 5A, panels 4 -6). To confirm this association, we performed fractionation experiments (Fig. 5B) using whole cell homogenate (lane 1) and cytoplasmic (lane 2) and mitochondrial fractions (lane 3) of N2A cells. The mitochondrial fraction was verified for purity by detection of the mitochondrial transmembrane voltage-dependent anion-selective channel/porin protein and the absence of actin (cytoplasmic component). Fig.  5B clearly shows that MARK staining occurs in the cytoplasmic fraction and in the mitochondrial fraction (lanes 2 and 3), suggesting a possible role of MARK2 in mitochondrial function. To characterize the linkage of MARK2, PINK1 FL , ⌬N-PINK1, and mitochondria, primary cortical neurons were transfected with adenoviral YFP-MARK2, PINK1 FL -YFP, or ⌬N-PINK1-YFP and stained for mitochondria. As mentioned above, high expression of MARK2 leads to phosphorylation of tau, breakdown of the microtubule network, and cell death. The adenoviral expression system was chosen because the protein expression is weak and in these conditions hardly toxic. Fig. 5C illustrates the distribution of mitochondria in neurites and cell bodies. In untransfected primary cortical neurons, the mitochondria are evenly spread out in axons and concentrated at axon roots in the cell body (Fig. 5C, panel 1). Neurons transfected with CFP-MARK2 or PINK1 FL -YFP by adenovirus show nearly the same mitochondrial distribution as untrans-fected cells (Fig. 5C, panels 2-5). Conspicuously, neurons transfected with PINK1 FL show the highest mitochondrial density in the initial part of axons (Fig. 5C, panel 5, arrow). Expression of ⌬N-PINK1 leads to a clear change of the distribution of mitochondria: they accumulate partially in neurites or are retained in the cell body (Fig. 5C, panels 6 and 7,  arrows), indicating a transport disorder. Taken together, these results show that PINK1 FL , ⌬N-PINK1, and MARK2 colocalize in part with mitochondria and potentially influence mitochondrial transport.
Cotransfection of MARK2 and PINK1 FL enhanced the concentration of mitochondria in the cell body (consistent with an inhibition of anterograde transport; see below). By contrast, mitochondria in neurons transfected with MARK2 and ⌬N-PINK1 show local accumulations in axons (Fig. 5C, panels  11-13, arrows). The transfection rate was low, and transfected neurons tend to degenerate. It seems that overexpression of both kinases is toxic for the cell, and transport of mitochondria is impaired.

Movement of Mitochondria Is Perturbed in MARK2/⌬N-PINK1-transfected Retinal Ganglion Cells-A proper mito-
chondrial localization is essential for neuronal function because of the high demand of energy, especially in active growth cones (38). Considering that MARK2 is involved in the regulation of microtubule-dependent transport (1) and that PINK1 is important for the anterograde transport of mitochondria along microtubules (36), we next investigated mitochondrial traffic in axons. Chicken RGCs were transfected with adenovirus encoding PINK1 FL or ⌬N-PINK1 and MARK2 WT either alone or in combination. The advantage of these cells is that the polarities of the axons are well defined so that transport directions are readily analyzed. 20 h after transfection, the neurons were stained with marker dye reporting on total mitochondria. Control neurons display a uniform distribution of mitochondria in axons similar to neurons transfected with MARK2 WT (Fig. 6A,  panels 1 and 2). Likewise, PINK1 FL is mainly colocalized with uniformly distributed mitochondria. By contrast, mitochondria in neurons transfected with ⌬N-PINK1 are mostly accumulated in axonal swellings characteristic of degeneration (Fig. 6A,  panel 5), indicating an inhibition of axonal transport. This effect was enhanced by cotransfection of ⌬N-PINK1 in combination with MARK2 WT (Fig. 6A, panel 6). Quantification of mitochondrial density revealed that the number of axonal mitochondria increases ϳ2-fold in ⌬N-PINK1-transfected cells (Fig. 6B, bar 5) in comparison with controls (Fig. 6B, bar 1) or MARK2 WT -transfected cells (Fig. 6B, bar 2) evaluated in the proximal 50 m of the axons. The mitochondrial density in axons of PINK1 FL -transfected cells dropped slightly (Fig. 6B,  bar 3), and this effect was further enhanced by MARK2 WT (Fig.  6B, bar 4). Cotransfection of ⌬N-PINK1 with MARK2 WT led to a much higher mitochondrial density compared with the effect of ⌬N-PINK1 alone (Fig. 6B, bars 5 and 6), supporting the hypothesis of a MARK2 WT /⌬N-PINK1-induced mitochondrial transport disorder.
Indeed, the change in distribution of mitochondria in axons corresponds to a change in mobility. We analyzed mitochondrial movement by live cell confocal microscopy. Untransfected cultures display the typical discontinuous saltatory movements of mitochondria with pauses often followed by a reversal of direction (Fig. 7, A, panel 1, and C, bar 1). In the case of MARK2 WT -, PINK1 FL ,-or ⌬N-PINK1-transfected cells, the saltatory movements remain mostly intact (Fig. 7A, panels 2, 3,  and 5). By contrast, in MARK2 WT -and ⌬N-PINK1-cotransfected cells, the movement of mitochondria was strongly inhibited. At 20 h after transfection, the mitochondria become mostly stationary (Fig. 7A, panel 6), whereas cotransfection of MARK2 WT and PINK1 FL had no influence on mitochondrial motion (Fig. 7A, panel 4).
For time periods of 10 -20 min, we recorded the parameters of speed, run length, direction, pausing time, and their dependence on PINK1 FL , ⌬N-PINK1, and MARK2 WT . Quantification showed that in control cells 20% of the stained particles are stationary (Fig. 7B, bar 1). The fraction of immobile mitochondria did not change significantly in PINK1 FL -transfected cells (Fig. 7B, bar 3) and increased to ϳ35% in cells transfected with MARK2 WT alone or PINK1 FL in combination with MARK2 WT (Fig. 7B, bars 2 and 4). A further increase of stationary mitochondria to ϳ45% can be observed in ⌬N-PINK1-transfected cells (Fig. 7B, bar 5). This effect is boosted to ϳ76% in ⌬N-PINK1 plus MARK2 WT -transfected cells (Fig. 7B, bar 6). Thus, ⌬N-PINK1 forces mitochondria to become stationary in cooperation with MARK2.
⌬N-PINK1 Promotes Anterograde and PINK1 FL Promotes Retrograde Mitochondrial Transport, Both Depending on MARK2-The next question was whether the moving fraction of mitochondria displays changes in their transport direction dependent on PINK1 FL , ⌬N-PINK1, and MARK2 transfection. This was tested in adenoviral transfected retinal ganglion cell axons. Overall in control cultures, most of the mitochondria moved anterogradely (Fig. 7C, bar 1; ϳ70%). By contrast, in MARK2 WT -or in PINK1 FL -transfected cells the antero-and retrograde mitochondrial transport rates were nearly balanced with a slight excess of retrograde events (Fig. 7C, bars 2 and 3). Cotransfection of MARK2 WT and PINK1 FL strongly enhanced retrograde events to nearly 80% (Fig. 7C, bar 4, red box). However, in contrast to PINK1 FL , ⌬N-PINK1 reversed these effects and returned the ratio to the wild-type prevalence of anterograde movements, especially in the presence of MARK2 WT (Fig. 7C, compare bars 1, 5, and 6). In summary, our data suggest that PINK1 FL , ⌬N-PINK1, and MARK2 WT regulate the mitochondrial transport. PINK1 FL pushes the balance toward retrograde movements, whereas ⌬N-PINK1 promotes anterograde movements; in both cases, MARK2 WT enhances these effects (illustrated by the anterograde/retrograde ratios in Fig.  7D, bars 4 and 6, red boxes). The enhancement of anterograde movements by ⌬N-PINK1 is consistent with the observed mitochondrial accumulations and axonal swellings (see Fig. 6A, panels 5 and 6). Mutation of MARK2 Target Site Thr-313 in ⌬N-PINK1 and PINK1 FL Is Toxic in Primary Cortical and CHO Cells-PINK1 Thr-313 is frequently mutated in PD, and it is the main phosphorylation site of the upstream kinase MARK2. To clarify the importance of the site, we created different PINK1 Thr-313 mutants and observed their subcellular localization as well as their effects on mitochondria in neuronal and CHO cells. Expression of PINK1 T313E or PINK1 T313M was toxic for HEK293 and CHO cells. It was therefore not possible to harvest a sufficient amount of mutated PINK1 adenovirus because most of the transfected cells died. Expression of ⌬N-PINK1 T313M in primary cortical neurons led to abnormal mitochondrial accumulations in the cell soma and retraction of axons and dendrites (Fig. 8A, panels 1-3, arrows). Similar effects were observed even when the cellular microtubules were stabilized by Taxol (Fig. 8B, panels 1-3, arrow). Expression of PINK1 FL/T313M in Taxol-treated CHO cells led to the degradation of mitochondria (Fig. 8B, panels 4 -6, arrow). This mutant is not inactive and can still be phosphorylated by MARK2 (see Fig. 4A), explaining its toxicity. The quantification of viable transfected cells confirms these results. The survival rate drops to 40% in the case of PINK1 FL/T313M and to 6% in the case of ⌬N-PINK1 T313M expression compared with the expression of the non-mutated form (see Fig. 8C). The T313E mutant was toxic to cells both for full-length and truncated PINK1 because this mutant showed a higher activity in combination with its activating kinase MARK2 (see Fig. 4, A and B).

DISCUSSION
This report describes the interaction of two protein kinases, MARK2 and PINK1, both of which are implicated in neurodegenerative diseases (Alzheimer disease and PD). One of the main known functions of MARK2 is to phosphorylate tau and detach it from microtubules. This is an early event in Alzheimer disease, suggesting that overactivation of MARK2 might play a role in the breakdown of microtubules and the interruption of axonal transport (39). Conversely, in neurons with tau elevated by transfection, MARK2 can rescue transport inhibition by phosphorylating and detaching tau from microtubules (1). To analyze the MARK2 signaling pathways in greater detail, we set out to identify new regulatory partners of MARK2. One conspicuous hit was PINK1, a mitochondrially targeted serine/ threonine kinase with neuroprotective features (9,10). Loss-offunction mutations in the PINK1 gene result in recessive heritable forms of parkinsonism (40). Recent studies have shown that PINK1 plays a crucial role in the regulation of mitochondrial dynamics, transport, and function (12,41). Upon entry into mitochondria, full-length PINK1 protein can be cleaved by voltage-sensitive proteolysis (42). The most stable cleavage product is the 45-kDa fragment, corresponding roughly to ⌬N-PINK1 (25). Cytoplasmic actions of ⌬N-PINK1 are critical for cell survival because expression of ⌬N-PINK1 protects neurons against the dopaminergic neurotoxin 1methyl-4-phenyl-1,2,3,6-tetrahydropyridin (23). However, the mechanism of this ⌬N-PINK1 function is unknown. A major limitation in understanding the different functions of this protein kinase with dual localization is the lack of reliable PINK1 antibodies and low levels of endogenous expression (43). Therefore, much of the existing data has been based on the expression of tagged, exogenous PINK1.
We found an interaction of MARK2 especially with ⌬N-PINK1 that is based on their catalytic domains (Fig. 1B). Consistent with this, primary cortical neurons show colocalization of overexpressed ⌬N-PINK1 and MARK2 in dotted structures in axons and dendrites identified as mitochondria (Figs. 2 and 5A). PINK1 FL is predominantly localized in the mitochondrial outer membrane with the kinase domain facing the cytoplasm (24), allowing binding and potentially further modification of PINK FL by MARK2.
Nevertheless, the questions arose why several cleavage fragments of PINK1 are found in different cell compartments and how they become active. Recent studies suggest that PINK1 FL is active in the PINK1/Parkin pathway. Parkin is a cytoplasmic E3 ubiquitin ligase and can be phosphorylated by PINK1 FL (44). Both proteins cooperate to control mitochondrial clearance, known as mitophagy (for reviews, see Refs. 38 and 45 and see Ref. 46). However, neither the exact molecular mechanism how PINK1 recruits Parkin nor the possible role of Parkin phosphorylation are known in detail.
The interaction of MARK2 with ⌬N-PINK1 and PINK1 FL has distinct consequences for their kinase activities in vitro. MARK2 phosphorylates and activates preferentially ⌬N-PINK1 (Figs. 3A and 4, A and B). In the case of PINK1 FL , its interaction with MARK2 during the phosphorylation reaction was too transient to be detectable by coimmunoprecipitation, but it could be observed with the non-phosphorylatable mutant PINK1 FL/T313M , which locks the two partners in an inactive complex (an analogous situation was observed previously for the interaction between MARK2 and PAK5 (6)). In addition, MARK2 induced the phosphorylation of PINK1 FL in transfected CHO cells (supplemental Fig. S1). Phosphopeptide mapping followed by mass spectrometry analysis revealed that MARK2-dependent phosphorylation of ⌬N-PINK1 occurs mainly at residue Thr-313, which lies in the kinase domain and possibly leads to a stabilization of the scaffold that fixes the Mg-ATP, a requirement for the activity of the kinase (Fig. 3C). The mutation T313M in PINK1 is associated with familial PD (26); although mutation of this site to alanine decreased the phosphorylation of ⌬N-PINK1 by MARK2, it was not completely abolished (Fig. 4A, lane/bar 9), and activation could still occur (Fig. 4B, lane/bar 9), probably via conformational changes and phosphorylation of additional sites (Fig. 3B). However, mutation of Thr-313 to glutamate further enhanced the phosphorylation and activation of ⌬N-PINK1 by MARK2 (Fig.  4, A, lane/bar 8, and B, lane/bar 8). Thr-313 could be a priming phosphorylation site, changing the conformation of the kinase and preparing it for further modifications. The importance of this PINK1 phosphorylation site is further emphasized by the  5, arrow). The data argue that MARK2 is an upstream regulator of ⌬N-PINK1 and suggest that failure of this cascade contributes to the development of PD. The mutation could also have effects on known PINK1 substrates like Omi/HtrA2 or TRAP-1 (22,30), but it is questionable whether these proteins are phosphorylated by ⌬N-PINK1 under physiological conditions because both are mitochondrial and not cytoplasmic.
Within neurons, endogenous PINK1 and MARK2 partly colocalize on mitochondria, especially in axons and dendrites (Fig. 5, A and C, and supplemental Fig. S2). Fractionation studies revealed that endogenous MARK is found in the mitochondrial fraction of N2a cells (Fig. 5B). From these findings, we hypothesize that MARK2 may be recruited to mitochondria via ⌬N-PINK1 and/or PINK1 FL . This would alter the mitochondrial distribution because MARK2 in combination with PINK1 changes mitochondrial transport parameters (Figs. 5-7). New aspects on the relationship between PINK1 and mitochondrial transport regulation have been published recently (47). PINK1 phosphorylates the GTPase Miro, a component of a kinesinadaptor complex, and thus induces its Parkin-dependent degradation. The resulting decrease in mitochondrial movement may represent a quality control mechanism of defective mitochondria (47). In our hands, particularly ⌬N-PINK1 induced an arrest of mitochondrial motility, which was even further enhanced by MARK2 coexpression (Fig. 7B). In addition, ⌬N-PINK1 increased anterograde movement and eventually led to enhanced mitochondrial density in axons. MARK2 reinforced this effect, possibly by phosphorylation and activation of ⌬N-PINK1. Consistent with this, a high membrane potential promotes the anterograde transport of mitochondria (48) and also promotes the proteolysis of full-length PINK1 into ⌬N-PINK1 (42). Conversely, retrograde transport is favored for mitochondria with low membrane potential destined for mitophagy (49); in this case, the cleavage of PINK1 FL is inhibited in agreement with the transport characteristics induced by PINK1 FL ϩ MARK2 (Fig. 7). Mounting evidence indicates that Parkin is recruited via PINK1 FL to defective mitochondria, inducing their degradation (for a review, see Ref. 50). Thus, Parkin-decorated mitochondria assemble as large clusters primarily in the lysosome-rich perinuclear area (42,51). This effect seems to be influenced by MARK2 (Fig. 7, C and D). MARK2 could phosphorylate PINK1 FL and consequently enhance the binding and possibly the phosphorylation of Parkin and Miro by PINK1 FL . This is also suggested by the results of a protein stability assay. MARK2 T208/S212A enhanced particularly the stability of PINK1 FL (Fig. 4D). Conversely, healthy mitochondria promote the cleavage of PINK1 FL , leading to high levels of ⌬N-PINK1, which is not able to translocate Parkin to impaired mitochondria (51). This suppresses mitophagy and could explain why mitochondria accumulate in the cell soma of ⌬N-PINK1 T313M -transfected cells (Fig. 8).
Mitochondria are transported along microtubules by motor proteins kinesin (forward) and dynein (reverse). The kinesinadaptor complex attached to the outer membrane comprises the GTPase Miro, kinesin heavy chain, and the adaptor protein Milton (52). PINK1 FL is also attached to this complex, and even ⌬N-PINK1 can be targeted to it despite the lack of the mitochondrial targeting sequence (24,36). These observations argue that ⌬N-PINK1 can participate in the control of anterograde mitochondrial trafficking and influence kinesin indirectly via Milton, consistent with the enhanced anterograde transport in cells overexpressing ⌬N-PINK1 (Fig. 7, C and D).