Identification of Neuronal Substrates Implicates Pak5 in Synaptic Vesicle Trafficking

Identification of neuronal substrates implicates Pak5 in synaptic vesicle trafficking

Todd I. Strochlica, Susanna Concilioa, Julien Viaudb, Ryan A. Eberwinea, Lisa Epstein Wongc, Audrey Mindenc, Benjamin E. Turkd, Markus Plomanne, and Jeffrey R. Petersona,1

aCancer Biology Program, Fox Chase Cancer Center, 333 Cottman Avenue, Philadelphia, PA 19111; bInstitut National de la Sante et de la Recherche Medicale U1048 Equipe B. Payrastre Bâtiment B, Pavillon Lefebvre, BP 3028, 31024 Toulouse, France; cSusan Lehman Cullman Laboratory for Cancer Research, Department of Chemical Biology, Ernest Mario School of Pharmacy, Rutgers University, 164 Frelinghuysen Road, Piscataway, NJ 08854; dDepartment of Pharmacology, Yale University School of Medicine, 333 Cedar Street, New Haven, CT 06520; and eCenter for Biochemistry, Medical Faculty of the University of Cologne, Joseph-Stelzmann-Strasse 52 50931 Cologne, Germany

Edited by Pietro De Camilli, Yale University and Howard Hughes Medical Institute, New Haven, CT, and approved January 17, 2012 (received for review October 7, 2011)

Synaptic transmission is mediated by a complex set of molecular in the brain (14, 15). Consistent with some level of functional re- events that must be coordinated in time and space. While many dundancy between these Pak isoforms, no phenotype has been proteins that function at the synapse have been identified, the reported for PAK5 or PAK6 single knockout mice (10, 16), but signaling pathways regulating these molecules are poorly under- PAK5/PAK6 double knockout mice exhibit defects in behavior, stood. Pak5 (p21-activated kinase 5) is a brain-specific isoform of memory, and learning (16). The molecular mechanisms underly- the group II Pak kinases whose substrates and roles within the ing these cognitive defects, however, have yet to be determined. central nervous system are largely unknown. To gain insight into To gain further insight into the functions of Pak5 in vivo, we the physiological roles of Pak5, we engineered a Pak5 mutant to identified and purified direct substrates of this kinase from mur- selectively radiolabel its substrates in murine brain extract. Using ine brain extract. Here we report the identification of two novel this approach, we identified two novel Pak5 substrates, Pacsin1 Pak5 substrates, Synaptojanin1 and Pacsin1, proteins that directly and Synaptojanin1, proteins that directly interact with one another bind to one another at the synapse to regulate vesicle dynamics.

to regulate synaptic vesicle endocytosis and recycling. Pacsin1 Furthermore, Pak5 phosphorylation promotes the interaction of BIOCHEMISTRY and Synaptojanin1 were phosphorylated by Pak5 and the other Synaptojanin1 with Pacsin1, implicating Pak5 in synaptic vesicle group II Paks in vitro, and Pak5 phosphorylation promoted Pacsin1- trafficking. Synaptojanin1 binding both in vitro and in vivo. These results implicate Pak5 in Pacsin1- and Synaptojanin1-mediated synaptic Results vesicle trafficking and may partially account for the cognitive and Identification of Pak5-Specific Substrates in Murine Brain. Gate- behavioral deficits observed in group II Pak-deficient mice. keeper residue mutants of several protein kinases have been employed as tools to identify direct kinase substrates (17). These Syndapin1 ∣ F-BAR ∣ autoinhibition ∣ kinase substrate mutations enlarge the ATP binding pocket to accommodate the binding of unnatural ATP analogues modified by bulky substitu- egulated signal transduction at neuronal synapses underlies ents at the N6 position that the majority of wild-type kinases can- the development and function of the nervous system and is not use. A gatekeeper residue mutant, Pak5-M523G, but not R 32 crucial for proper neurotransmission. This is highlighted by a wild-type Pak5, can utilize [ P]-N6-methylbenzyl-ATP to phos- number of neurological disorders such as mental retardation and phorylate Pak1 (18) and myelin basic protein (MBP) (Fig. S1). autism that result from defects in synaptic signaling cascades (1). Likewise, we found that the analogous Pak2 mutant (M323G) 32 In particular, aberrant signaling by protein kinases functioning can also utilize [ P]-N6-methylbenzyl ATP as a substrate whereas downstream of Rho GTPases has been linked to these disorders wild-type Pak2 cannot (Fig. S1). We used these Pak2/5 mutants to (2, 3). Inactivation of both the RhoA/Rho-kinase (ROCK) path- identify direct and specific substrates of Pak5 in fractionated way and the Rac/Cdc42/p21-activated kinase (Pak) pathway have mouse brain lysates (Fig. S2). been implicated in the development of mental retardation Brain extract was fractionated by cation exchange chromato- through dysregulation of neurotransmitter receptor endocytosis graphy and a sample of each fraction was incubated with either 32 and trafficking (4–6). Loss of Pak3 expression as well as specific [ P]-N6-methylbenzyl-ATP alone or together with Pak5-M523G. mutations in PAK3 are associated with abnormalities in synaptic Reactions were analyzed by Coomassie blue-stained SDS/PAGE plasticity and cognition (6, 7) and a decreased density of dendritic and autoradiography to identify fractions containing proteins spines (8). In addition, both Pak1 and Pak3 are involved in phosphorylated in a Pak5-dependent manner. A Pak5 substrate regulating brain mass through modulation of neuronal size and of 150 kDa and a second substrate of 55 kDa were identified, synaptic complexity (9). These findings underscore the need to and fractions containing these proteins were pooled and further further characterize the roles of protein kinases, particularly fractionated by anion exchange chromatography. Individual frac- those of the Pak family, in neuronal physiology and pathology. tions containing p150 and p55 were identified by Pak5-M523G Pak5 is a brain-specific serine/threonine protein kinase ex- kinase assays. In parallel, each fraction was also subjected to pressed exclusively in neurons but not glial cells (10) and whose kinase assays using Pak2-M323G. Both p150 and p55 were spe- functions within the central nervous system are poorly described (11). Six Pak isoforms exist in mammals, classified into two Author contributions: T.I.S., J.V., and J.R.P. designed research; T.I.S., S.C., J.V., R.A.E., B.E.T., groups based on sequence homology and regulatory properties. and M.P. performed research; L.E.W., A.M., and M.P. contributed new reagents/ The roles of the group I Paks (Pak1–3) in signaling pathways that analytic tools; T.I.S., S.C., J.V., R.A.E., and B.E.T. analyzed data; and T.I.S. and J.R.P. wrote mediate cell motility, proliferation, and survival are well docu- the paper. mented; however, the functions of the group II Paks (Pak4–6) The authors declare no conflict of interest. are much less well understood (12). This article is a PNAS Direct Submission. Among the group II Paks, only Pak4 is ubiquitously expressed 1To whom correspondence should be addressed. E-mail: [email protected]. and is essential for viability (13), whereas Pak5 and Pak6 have a This article contains supporting information online at www.pnas.org/lookup/suppl/ more limited tissue expression pattern with especially high levels doi:10.1073/pnas.1116560109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1116560109 PNAS Early Edition ∣ 1of6 Downloaded by guest on September 26, 2021 Fig. 1. Identification of two Pak5 specific substrates in murine brain. Autoradiograms and corresponding Coomassie-blue-stained gels of pools 1 and 2. White arrowhead highlights the 55-kDa Pak5-specific substrate in pool 1. Black arrowhead highlights the 150-kDa Pak5-specific substrate in pool 2.

cifically phosphorylated by Pak5 but not Pak2 (Fig. 1). Coomassie with a Synaptojanin1 antibody. Western blotting with the same anti- blue-stained bands that correlated by molecular weight and by body demonstrated depletion of this band from the fraction and its cofractionation with the [32P]-labeled substrates were excised and appearance in the immunoprecipitate (Fig. 2A, Upper). Samples of subjected to identification by MS/MS mass spectrometry of tryptic the starting fraction, the immunodepleted fraction, and the immu- peptides. Seventy-three peptides corresponding to 55% of the ami- noprecipitated material were then used in an in vitro kinase assay no acid sequence of Synaptojanin1 were identified in the 150 kDa with recombinant Pak5 (Fig. 2A, Lower). A 150-kDa band was radi- band, representing more total peptide counts and greater percent olabeled only when Pak5 was present, did not appear in the kinase coverage than any other protein identified in this band (Fig. S3A). assay using the immunodepleted fraction, and was present in the Similarly, the protein with the most peptide counts (36 peptides) anti-Synaptojanin1 immunoprecipitate. Taken together, these re- and greatest percent coverage (64%) in the 55 kDa band was Pac- sults demonstrate that the 150-kDa Pak5 substrate is Synaptojanin1. sin1 (Fig. S3B). Importantly, Synaptojanin1 and Pacsin1 have cal- Synaptojanin1 consists of an N-terminal Sac1 phosphatase culated molecular weights of 145 kDa and 52 kDa, respectively, domain, a central 5- phosphatase domain, and a C-terminal consistent with the observed mobility of the Pak5 substrates by proline-rich domain (Fig. 2B). To determine which domain(s) SDS/PAGE. of Synaptojanin1 are phosphorylated by Pak5, full-length Synap- tojanin1 or truncation constructs (Fig. 2B) were expressed, immu- Synaptojanin1 and Pacsin1 Are Novel Pak5 Substrates. Synaptojanin1 nopurified, and used in in vitro kinase assays with Pak5 (Fig. 2C). is a member of the phosphoinositide 5-phosphate family of phos- Synaptonjanin1 is phosphorylated only in constructs containing phatases. This enzyme dephosphorylates phosphatidylinositol 4,5- the proline-rich domain, indicating that the phosphorylation bisphosphate (PIP2), a prerequisite for the uncoating and recycling site(s) likely resides within this region. of synaptic vesicles (19). As a consequence, Synaptojanin1−∕− mice To predict the specific phosphorylation site within this domain, display elevated steady-state levels of PIP2 with an accumulation of we elucidated the amino acid sequence selectivity of Pak5 using a clathrin-coated vesicles in nerve terminals, and these animals die positional scanning peptide library approach (Fig. S4). This meth- shortly after birth (20). Toconfirm the identity of the 150-kDa band od uses partially degenerate peptide mixtures to systematically as Synaptojanin1, we immunodepleted a p150-containing fraction test all of the naturally occurring amino acids at each of nine

Fig. 2. The p150 Pak5 substrate is the phosphoinositide phosphatase Synaptojanin1. (A) Immunodepletion and kinase assays using a p150-containing fraction. Fraction 17 (Fig. 1, pool 2) was immunodepleted with an anti-Synaptojanin1 antibody and the starting fraction, depleted fraction, and immunoprecipitated material were analyzed by Western blotting with an anti-Synaptojanin1 antibody (Top) and by Pak5 in vitro kinase assays (Bottom). (B) Synaptojanin1 domain architecture and truncation/point mutants tested for Pak5 phosphorylation. All constructs were myc-tagged at the C terminus. (C) Mapping of the Pak5 phosphorylation site in Synaptojanin1. The indicated constructs were expressed in and immunoprecipitated from HEK293 cells and used in Pak5 in vitro kinase assays. Reactions were analyzed by Coomassie-blue-stained SDS/PAGE and autoradiography. Asterisks denote bands corresponding to Synaptojanin1 proteins.

2of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1116560109 Strochlic et al. Downloaded by guest on September 26, 2021 positions surrounding a phosphoacceptor site for phosphoryla- Three isoforms of Pacsin are expressed in higher eukaryotes. tion by a given kinase (21). Overall, we found that the peptide Pacsin1 is brain-specific whereas Pacsin2 and Pacsin3 are more substrate specificity for Pak5 was indistinguishable from the pre- widely expressed (27). An alignment of the sequences of Pacsin1/ viously reported specificity of Pak4 (22), and like all other Pak 2/3 reveals that serine 343 within the Pak family consensus motif isoforms, the most important determinant for Pak5 phosphoryla- is not conserved in Pacsin2 and 3 (Fig. S5A). Consistent with this tion is the presence of an arginine at the -2 position. However, finding, Pacsin1 is phosphorylated in vitro by Pak5 whereas unlike the group I isoforms Pak1 and Pak2 (22), the second stron- Pacsin2 and 3 are not (Fig. S5B), demonstrating that only the gest positive selection by Pak5 was for serine at the +3 position. brain-specific Pacsin1 isoform is a substrate of Pak5. The Pak5 selectivity data was used to quantitatively rank all possible sites for Pak5 phosphorylation within the proline-rich Pacsin1 and Synaptojanin1 are Group II Pak-Specific Substrates. domain of Synaptojanin1, and serine 1291 ranked as the most Although the kinase domains of the group I and group II Paks likely site by this analysis. When we mutated serine 1291 to ala- are quite divergent, the majority of group II Pak substrates are nine (S1291A) within the context of the full-length protein, it was also phosphorylated by the group I Paks (12). Our data indicate no longer phosphorylated, indicating that Pak5 phosphorylates that Pacsin1 and Synaptojanin1 are phosphorylated specifically Synaptojanin1 exclusively on serine 1291. by Pak5 but not by Pak2 (Fig. 1), suggesting that these proteins Pacsin1, also known as Syndapin1, is an evolutionarily con- may be group II-specific Pak substrates. To directly test this, we served, brain-specific F-BAR (Fer-Cip4-Bin/Amphiphysin/Rvs) performed in vitro kinase assays with all six Pak isoforms using and SH3 (Src homology 3)-domain-containing protein (23, 24). GST-Pacsin1 or Synaptojanin1-myc and, as a control, the generic The SH3 domain of Pacsin1 binds to the proline-rich regions of Pak substrate MBP. Whereas MBP is phosphorylated by all Pak several other proteins including dynamin1, N-WASP (neuronal isoforms (Fig. 4 A and B, Lower Panels), Pacsin1 and Synaptoja- Wiskott–Aldrich syndrome protein), and Synaptojanin1 (25) nin1 are preferentially phosphorylated by all three of the group II while the F-BAR domain is implicated in sensing and/or inducing Paks (Fig. 4 A and B, Upper Panels, lanes 5–7, and Fig. 4C) but not membrane curvature (26). Pacsin1 was immunoprecipitated from by the group I Paks (Fig. 4 A and B, Upper Panels, lanes 1, 2, and 4, a p55-containing fraction, and the starting fraction, the immuno- and Fig. 4C). depleted fraction, and the beads from the immunoprecipitation Two amino acids have been shown to play a critical role in dic- were analyzed by Western blotting to confirm Pacsin1 depletion tating the substrate selectivity of all Paks by influencing the pre- (Fig. 3A, Top). Samples of the starting fraction, the immunode- ferences of these kinases for particular residues at the +2 and +3

pleted fraction, and the immunoprecipitate were then used in positions of a substrate (22). Exchange of these residues partially BIOCHEMISTRY Pak5 kinase assays (Fig. 3A, Lower). A 55-kDa band was strongly swaps the peptide substrate specificity of group I to that of group radiolabeled by Pak5 in Pacsin1 immunoprecipitates, demon- II and vice versa (22). Therefore, we asked whether the specifi- strating that Pacsin1 is the 55-kDa substrate of Pak5. city-swapping mutant of Pak2 (P286Q, K287R; Pak2-QR) could Pacsin1 contains an N-terminal F-BAR domain and a C-term- phosphorylate Pacsin1 and Synaptojanin1 in vitro. The results in- inal SH3 domain connected by an unstructured linker region dicate that Pak2-QR was able to phosphorylate both substrates (Fig. 3B). To determine which domains of Pacsin1 are phosphory- whereas wild-type Pak2 could not (Fig. 4 A and B, Upper Panels, lated by Pak5, we expressed and purified GST-fusion proteins lanes 2 and 3, and Fig. 4C), suggesting that recognition of the consisting of various regions of Pacsin1 (Fig. 3B). Only Pacsin1 motif surrounding the phosphoacceptor residue in these proteins constructs containing the linker region were phosphorylated by confers group II Pak selectivity. To directly test this, we generated Pak5 (Fig. 3C), demonstrating that Pak5 likely phosphorylates an 11 amino acid GST-fusion peptide consisting of the residues a site within this region. Pak5 consensus sequence analysis ranked immediately flanking the Pak5 phosphorylation site in Pacsin1 serine 343 as the most likely candidate, and consequently, its mu- (GST-AGDRGSVSSYD) and used this fusion peptide as a sub- tation to alanine (S343A) within the context of the full-length strate in in vitro kinase assays with Pak2 or Pak5. Whereas both protein abolished phosphorylation by Pak5 (Fig. 3C). These re- Pak2 and Pak5 robustly phosphorylated MBP (Fig. S6, lanes 3 sults indicate that Pak5 phosphorylates Pacsin1 exclusively on and 4), this short peptide was strongly phosphorylated by Pak5 serine 343. but to a much lesser degree by Pak2 (Fig. S6, lanes 7 and 8), de-

Fig. 3. The p55 Pak5 substrate is the F-BAR family member Pacsin1/Syndapin1. (A) Immunodepletion and kinase assays using a p55-containing fraction. Frac- tion 25 (Fig. 1, pool 1) was immunodepleted with an anti-Pacsin1 antibody and the starting fraction, depleted fraction, and immunoprecipitated material were analyzed by Western blotting with an anti-Pacsin1 antibody (Top) and by Pak5 in vitro kinase assays (Bottom). (B) Pacsin1 domain architecture and truncation/ point mutants analyzed for Pak5 phosphorylation. The indicated constructs were expressed and purified as GST-fusion proteins from E. coli. (C) Mapping of the Pak5 phosphorylation site in Pacsin1. Pak5 kinase assays were performed with the indicated GST-Pacsin1 fusion proteins as substrates. Reactions were analyzed by Coomassie-blue-stained SDS/PAGE and autoradiography. Asterisks denote bands corresponding to Pacsin1. Note that the GST-SH3 domain fusion comigrates with Pak5.

Strochlic et al. PNAS Early Edition ∣ 3of6 Downloaded by guest on September 26, 2021 Phosphorylation by Group II Paks Regulates Pacsin1-Synaptojanin1 Binding. The SH3 domain of Pacsin1 binds to the proline-rich domain of Synaptojanin1 (25). The identification of both Pacsin1 and Synaptojanin1 as Pak5 substrates and the subsequent mapping of phosphorylation sites to, or adjacent to, regions in these proteins involved in their association (Figs. 2C and 3C) led us to hypothesize that Pak5 phosphorylation may affect their binding. To test this, we transfected HEK293 cells with constructs encoding Synaptojanin1-myc and GFP-Pacsin1 with or without a plasmid encoding HA-Pak5. Synaptojanin1 was immunoprecipitated from cell lysates and the extent of Pacsin1 coprecipitation was assessed by immunoblotting with anti-GFP antibodies (Fig. 6A). The results indicate that a greater amount of Pacsin1 coprecipitated with Synaptojanin1 when Pak5 is expressed (compare lanes 3and5inFig.6A), implying that phosphorylation by Pak5 in- creases their association. We tested this via an alternative approach by cotransfection of epitope-tagged phosphomimetic forms of both Synaptojanin1 (S1291D) and Pacsin1 (S343E). Consistent with the previous result, the phosphomimetic forms (Fig. 6A, lane 4) showed increased binding compared to the wild-type proteins (Fig. 6A, lane 3). Taken together, these two independent approaches demonstrate that Pak5 phosphorylation promotes the interaction of Pacsin1 and Synaptojanin1. We further probed the individual contribution of each phos- phosite on Pacsin1-Synaptojanin1 binding by coimmunoprecipitation from cells expressing both wild-type proteins, both phosphomimetic mutants, or one phosphomimetic mutant with the other wild-type protein. While expression of either phosphomimetic enhances Pacsin1-Synaptojanin1 binding (Fig. 6B, lanes 2 and 3), greatest binding is detected by coexpression of both phosphomimetic forms (Fig. 6B, lane 4). These results suggest that both sites of phosphorylation (S343 in Pacsin1 and S1291 in Synaptojanin1) contribute to the Pacsin1-Synaptojanin1 interaction. Finally, we investigated the physiological relevance of group II Pak phosphorylation on Pacsin1-Synaptojanin1 binding in the brain. Endogenous Synaptojanin1 was immunoprecipitated from Fig. 4. Pacsin1 and Synaptojanin1 are preferentially phosphorylated in vitro brain lysates of Pak5 and Pak6 single knockout and Pak5/Pak6 by group II Paks. (A and B) Pacsin1 and Synaptojanin1 are group II Pak-specific double knockout mice. Substantially less Pacsin1 coprecipitated substrates. GST-Pacsin1 (Fig. 4A, Upper Panels), Synaptojanin1-myc (Fig. 4B, with Synaptojanin1 from brains of Pak5/Pak6 double knockout Upper Panels) and MBP (Lower Panels) were used as substrates in in vitro C kinase assays with Paks1-6 and Pak2-P286Q/K287R (Pak2-QR). Reactions were mice compared to the single knockouts (Fig. 6 ). Collectively, analyzed by silver-stained or Coomassie-blue-stained SDS/PAGE, as indicated, these results indicate that phosphorylation by group II Paks reg- and autoradiography. * indicates autophosphorylation of the corresponding ulates the interaction between Pacsin1 and Synaptojanin1 both in Pak isoform. (C) Bands were quantified and plotted as the ratio of Pacsin1 or vitro and in vivo. Synaptojanin1 phosphorylation to MBP phosphorylation. Gray bars denote Paks with group I specificity and black bars denote Paks with group II speci- Discussion ficity. Among the six p21-activated kinases, Pak5 is one of the least char- acterized and most poorly understood isoforms. Previous reports monstrating that these residues form a specific recognition site for have described a role for Pak5 in antiapoptotic signaling (29) (30), group II Paks. and overexpression of Pak5 results in increased neurite outgrowth in N1E-115 neuroblastoma cells (31), although the mechanism by Pacsin1 Is Phosphorylated in Vivo by Group II Paks. To determine the physiological relevance of Pak5 phosphorylation, we developed a phosphospecific antibody against Pacsin1-S343 (Fig. S7A). A single band of 52 kDa was detected by Western blotting using the antibody, indicating that Pacsin1 is phosphorylated in vivo on serine 343 (Fig. S7B). Corroborating this result, phosphorylation of serine 343 in Pacsin1 has been detected by a large-scale mass spectrometry-based analysis of the mouse phosphoproteome (28). To test if phosphorylation of serine 343 was dependent on group II Paks, brain lysates were prepared from Pak5 and Pak6 single and double knockout mice and were immunoblotted with a total Pacsin1 antibody and with the phospho-Pacsin1-S343 antibody. Brain lysates from Pak5/Pak6 double knockout mice re- vealed substantially decreased levels of phosphorylated Pacsin1 compared to the single knockouts (Fig. 5) with the residual phos- Fig. 5. Pacsin1-S343 is phosphorylated in vivo in a Pak5/Pak6-dependent phorylation likely due to the presence of Pak4, which is also ex- manner. Brain lysates from mice of the indicated genotypes were immuno- pressed in the brain (13). Taken together, these data indicate that blotted with a phospho-Pacsin1-S343 antibody and with a total Pacsin1 anti- Pacsin1 is phosphorylated by Pak5 and Pak6 in vivo. body.

4of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1116560109 Strochlic et al. Downloaded by guest on September 26, 2021 which Pak5 mediates this effect is unknown. Several other Pak5 substrates have been identified including p120-catenin (32), Raf-1 (33), and Par-1 (34), but given the widespread expression of these proteins and the brain-restricted expression of Pak5, the physiological implications of these modifications is unclear. In this study, we demonstrate that Pak5 directly phosphorylates Pacsin1 and Synaptojanin1 in brain cytosol enhancing their association and possibly linking the activities of these two proteins both spatially and temporally. In neuromorphogenesis, Pacsin1 functions at the interface of endocytosis and membrane trafficking, acting as an adaptor to link membrane deformation via its F-BAR domain to vesicle internalization and trafficking via its SH3 domain (24, 35). Also functioning in neurons, the phosphatase activity of Synaptojanin1 is crucial for phosphoinositide homeostasis and for the maintenance of a functional pool of synaptic vesicles (36). Our data illustrating a role for Pak5 in mediating the interaction between Pacsin1 and Synaptojanin1 connect aspects of endocytosis and membrane recycling at the synapse and provide insight into how the functions of Pacsin1 and Synaptojanin1 may be coordinated in vivo. Recent structural studies have determined that Pacsin1 exists in an autoinhibited conformation in which the SH3 domain is bound to the F-BAR domain within the same molecule. This autoinhibited conformation is incompatible with SH3 domain- ligand binding and inhibits membrane tubulation by the F-BAR domain (37). We propose a model in which phosphorylation of Pacsin1 at serine 343 by Pak5 relieves this autoinhibition, allow- ing the F-BAR domain to generate membrane curvature and BIOCHEMISTRY the SH3 domain to bind Synaptojanin1 (Fig. 7). We further speculate that serine 1291 phosphorylation induces a conformational change in Synaptojanin1 that enhances binding to the SH3 domain of Pacsin1. In support of this model, phosphoregulation of F-BAR- and SH3-domain-containing proteins is emerging as a common mechanism to modulate the conformation, subcellular localization, and activity of proteins that contain these domains (38). A strik- ing difference between these proteins and Pacsin1, however, is that phosphorylation inhibits their association with known ligands while phosphorylation of Pacsin1 appears to promote its interaction with a known binding partner. A possible explanation for this discrepancy may be the fact that the linker regions connecting the F-BAR and SH3 domains in all of these proteins are phosphorylated at multiple sites. We propose that these linker regions func-

Fig. 6. Phosphorylation by group II Paks regulates Pacsin1-Synaptojanin1 binding. (A and B) Phosphorylation of Pacsin1-S343 and Synaptojanin1-S1291 enhances Pacsin1-Synaptojanin1 binding in vitro. HEK293 cells were transfected with the indicated constructs. The amount of GFP-Pacsin1 or GFP-Pac- sin1-S343E coprecipitating with Synaptojanin1 or Synaptojanin1-S1291D was assessed by immunoblotting with anti-GFP antibodies. Immunoblotting of whole cell lysates demonstrates equal expression of GFP-Pacsin1/GFP- Pacsin1-S343E and detection of HA-Pak5 and phospho-Pacsin1-S343 in the relevant samples. Bands from B (Middle) were quantified and normalized to the amount of wild-type Pacsin1 coprecipitating with wild-type Synapto- Fig. 7. Model: group II Paks regulate the F-BAR- and SH3 domain-mediated janin1 (lane 1). Means from three independent experiments are displayed. functions of Pacsin1. Phosphorylation by group II Paks (e.g., Pak5) and/or Error bars denote standard deviation. (C) Phosphorylation by group II Paks binding of phosphorylated Synaptojanin1 relieves intrinsic Pacsin1 autoinhi- regulates Pacsin1-Synaptojanin1 binding in vivo. Endogenous Synaptojanin1 bition by inducing a conformational change in the linker region. This allows was immunoprecipitated from brain lysates of mice of the indicated geno- the F-BAR domain of Pacsin1 to interact with and deform the plasma mem- types (Top). The amount of coprecipitating endogenous Pacsin1 (Middle) was brane and may enhance binding of ligands (i.e., additional molecules of determined by immunoblotting along with levels of total Pacsin1 (Bottom). Synaptojanin1) to the SH3 domain. A surface representation of the dimeric Bands were quantified and are displayed as the mean from three indepen- F-BAR domain (green) and the SH3 domain (blue) of Pacsin1 were generated dent experiments. Error bars denote standard deviation, and * indicates p using Pymol (Schrodinger, LLC) based on the crystal structure (37) (PDB ID value <0.01 for the double knockout compared to either single knockout. code 2X3X). The linker region was disordered in the crystal and is represented A:U: ¼ arbitrary units. here by a black line.

Strochlic et al. PNAS Early Edition ∣ 5of6 Downloaded by guest on September 26, 2021 tion as platforms for signal integration from a number of kinases The Pak5 phosphorylation sites and key elements of the Pak and phosphatases that ultimately determines whether the overall consensus motif in both Synaptojanin1 and Pacsin1 are evolutio- effect on protein function is stimulatory or inhibitory. narily conserved from jawed fish to mammals (Fig. S8), suggesting Within the central nervous system, the strength and maturity of important biological functions for these modifications. Interest- synaptic connections is defined by the number and types of neu- ingly, the presence of these proteomic features coincides with the rotransmitter receptors at the postsynaptic membrane. Pacsin1 acquisition of myelinated nerve fibers (43). It is therefore tempting binds to NR3A-containing NMDA receptors and promotes their endocytosis (39), and Synaptojanin1 has also been implicated in to speculate that myelination and Pak5-mediated phosphorylation endocytosis (40), specifically in postsynaptic AMPA receptor of Pacsin1 and Synaptojanin1 coevolved, as Pak5-induced associa- downregulation (41). Pak5−∕−∕Pak6−∕− mice exhibit defects in tion of Synaptojanin1 and Pacsin1 may enhance synaptic vesicle locomotion, learning, and memory (16), and based on our results, recycling and, like myelination, may facilitate neurotransmission. we propose that the cognitive and behavioral deficits observed in these mice may be partly attributable to altered endocytosis and Materials and Methods vesicle trafficking. Learning and memory, in particular, are depen- Pak5 Substrate Identification. Wild-type murine brain extract was fractio- dent on a form of synaptic plasticity known as long-term potentia- nated by cation and anion exchange chromatography, and samples of each tion, a process that relies heavily on the density of neurotransmitter fraction were subjected to in vitro kinase assays. Additional details are avail- receptors within the postsynaptic membrane (42). Defects in Sy- able as supplemental information. Experimental details regarding antibodies, plasmid construction, recombi- naptojanin1-Pacsin1-linked synaptic vesicle dynamics may perturb nant protein expression and purification, generation of [32P]-N6-methylbenzyl the number of NMDA- and AMPA-type receptors at the plasma ATP, preparation of mouse brain extract, chromatography, characterization of membrane with a subsequent inhibition of long-term potentiation. Pak5 substrate specificity determinants, in vitro kinase assays, cell culture and Our data indicate that Pacsin1 and Synaptojanin1 are phos- transfection, and coimmunoprecipitation assays are available as supplemental phorylated by the group II Paks but not by the group I Paks, repre- information. senting, to our knowledge, Pak substrates that are phosphorylated solely by the group II isoforms. Furthermore, our results suggest ACKNOWLEDGMENTS. We thank P. De Camilli, S. Knapp, J. Chernoff, M. Kelly, that the amino acid context surrounding the Pacsin1 phosphoryla- J. Fukui, L. O’Donnell, G. Rall, and W. Xu for reagents and discussions. This tion site dictates a strong preference for recognition by the group II work was supported by a W. W. Smith Charitable Trust Award to J.R.P., by an Paks, implying that the peptide substrate specificity differences American Cancer Society postdoctoral fellowship (PF-11-068-01-TBE) to T.I.S., identified in vitro for group I vs. group II Paks (22) are physiolo- and by National Institutes of Health Grants R01 GM083025 to J.R.P., T32 gically relevant. CA009035 to T.I.S., and P30 CA006927 to Fox Chase Cancer Center.

1. Boda B, Dubos A, Muller D (2010) Signaling mechanisms regulating synapse formation 23. Plomann M, et al. (1998) PACSIN, a brain protein that is upregulated upon differentia- and function in mental retardation. Curr Opin Neurobiol 20:519–527. tion into neuronal cells. Eur J Biochem 256:201–211. 2. Nadif Kasri N, Van Aelst L (2008) Rho-linked genes and neurological disorders. Pflugers 24. Kessels MM, Qualmann B (2004) The syndapin protein family: Linking membrane traf- Arch 455:787–797. ficking with the cytoskeleton. J Cell Sci 117:3077–3086. 3. Kreis P, Barnier JV (2009) PAK signalling in neuronal physiology. Cell Signal 21:384–393. 25. Qualmann B, Roos J, DiGregorio PJ, Kelly RB (1999) Syndapin I a synaptic dynamin- 4. Nadif Kasri N, Nakano-Kobayashi A, Malinow R, Li B, Van Aelst L (2009) The Rho-linked binding protein that associates with the neural Wiskott-Aldrich syndrome protein. mental retardation protein oligophrenin-1 controls synapse maturation and plasticity Mol Biol Cell 10:501–513. by stabilizing AMPA receptors. Genes Dev 23:1289–1302. 26. Wang Q, et al. (2009) Molecular mechanism of membrane constriction and tubu- 5. Rex CS, et al. (2009) Different Rho GTPase-dependent signaling pathways initiate se- lation mediated by the F-BAR protein Pacsin/Syndapin. Proc Natl Acad Sci USA quential steps in the consolidation of long-term potentiation. J Cell Biol 186:85–97. 106:12700–12705. 6. Boda B, et al. (2004) The mental retardation protein PAK3 contributes to synapse 27. Modregger J, Ritter B, Witter B, Paulsson M, Plomann M (2000) All three PACSIN iso- – formation and plasticity in hippocampus. J Neurosci 24:10816–10825. forms bind to endocytic proteins and inhibit endocytosis. J Cell Sci 113:4511 4521. 7. Meng J, Meng Y, Hanna A, Janus C, Jia Z (2005) Abnormal long-lasting synaptic plas- 28. Huttlin EL, et al. (2010) A tissue-specific atlas of mouse protein phosphorylation and – ticity and cognition in mice lacking the mental retardation gene Pak3. J Neurosci expression. Cell 143:1174 1189. 25:6641–6650. 29. Cotteret S, Chernoff J (2006) Nucleocytoplasmic shuttling of Pak5 regulates its anti- – 8. Kreis P, et al. (2007) The p21-activated kinase 3 implicated in mental retardation apoptotic properties. Mol Cell Biol 26:3215 3230. regulates spine morphogenesis through a Cdc42-dependent pathway. J Biol Chem 30. Cotteret S, Jaffer ZM, Beeser A, Chernoff J (2003) p21-Activated kinase 5 (Pak5) loca- lizes to mitochondria and inhibits apoptosis by phosphorylating BAD. Mol Cell Biol 282:21497–21506. 23:5526–5539. 9. Huang W, et al. (2011) p21-Activated kinases 1 and 3 control brain size through 31. Dan C, Nath N, Liberto M, Minden A (2002) PAK5, a new brain-specific kinase, coordinating neuronal complexity and synaptic properties. Mol Cell Biol 31:388–403. promotes neurite outgrowth in N1E-115 cells. Mol Cell Biol 22:567–577. 10. Li X, Minden A (2003) Targeted disruption of the gene for the PAK5 kinase in mice. Mol 32. Wong LE, Reynolds AB, Dissanayaka NT, Minden A (2010) p120-catenin is a binding Cell Biol 23:7134–7142. partner and substrate for group B Pak kinases. J Cell Biochem 110:1244–1254. 11. Wells CM, Jones GE (2010) The emerging importance of group II PAKs. Biochem J 33. Wu X, Carr HS, Dan I, Ruvolo PP, Frost JA (2008) p21 activated kinase 5 activates Raf-1 425:465–473. and targets it to mitochondria. J Cell Biochem 105:167–175. 12. Arias-Romero LE, Chernoff J (2008) A tale of two Paks. Biol Cell 100:97–108. 34. Matenia D, et al. (2005) PAK5 kinase is an inhibitor of MARK/Par-1, which leads to 13. Qu J, et al. (2003) PAK4 kinase is essential for embryonic viability and for proper neu- stable microtubules and dynamic actin. Mol Biol Cell 16:4410–4422. ronal development. Mol Cell Biol 23:7122–7133. 35. Dharmalingam E, et al. (2009) F-BAR proteins of the syndapin family shape the plasma 14. Pandey A, et al. (2002) Cloning and characterization of PAK5, a novel member of membrane and are crucial for neuromorphogenesis. J Neurosci 29:13315–13327. mammalian p21-activated kinase-II subfamily that is predominantly expressed in 36. Kim WT, et al. (2002) Delayed reentry of recycling vesicles into the fusion-competent – brain. Oncogene 21:3939 3948. synaptic vesicle pool in synaptojanin 1 knockout mice. Proc Natl Acad Sci USA 15. Lee SR, et al. (2002) AR and ER interaction with a p21-activated kinase (PAK6). Mol 99:17143–17148. – Endocrinol 16:85 99. 37. Rao Y, et al. (2010) Molecular basis for SH3 domain regulation of F-BAR-mediated 16. Nekrasova T, Jobes ML, Ting JH, Wagner GC, Minden A (2008) Targeted disruption of membrane deformation. Proc Natl Acad Sci USA 107:8213–8218. the Pak5 and Pak6 genes in mice leads to deficits in learning and locomotion. Dev Biol 38. Roberts-Galbraith RH, Gould KL (2010) Setting the F-BAR: functions and regulation of – 322:95 108. the F-BAR protein family. Cell Cycle 9:4091–4097. 17. Elphick LM, Lee SE, Gouverneur V, Mann DJ (2007) Using chemical genetics and ATP 39. Perez-Otano I, et al. (2006) Endocytosis and synaptic removal of NR3A-containing analogues to dissect protein kinase function. ACS Chem Biol 2:299–314. NMDA receptors by PACSIN1/syndapin1. Nat Neurosci 9:611–621. 18. Deacon SW, et al. (2008) An isoform-selective, small-molecule inhibitor targets the 40. Mani M, et al. (2007) The dual phosphatase activity of synaptojanin1 is required autoregulatory mechanism of p21-activated kinase. Chem Biol 15:322–331. for both efficient synaptic vesicle endocytosis and reavailability at nerve terminals. 19. McPherson PS, et al. (1996) A presynaptic inositol-5-phosphatase. Nature 379:353–357. Neuron 56:1004–1018. 20. Cremona O, et al. (1999) Essential role of phosphoinositide metabolism in synaptic 41. Gong LW, De Camilli P (2008) Regulation of postsynaptic AMPA responses by synap- vesicle recycling. Cell 99:179–188. tojanin 1. Proc Natl Acad Sci USA 105:17561–17566. 21. Hutti JE, et al. (2004) A rapid method for determining protein kinase phosphorylation 42. Lee HK, Kirkwood A (2011) AMPA receptor regulation during synaptic plasticity in hip- specificity. Nat Methods 1:27–29. pocampus and neocortex. Semin Cell Dev Biol 5:514–520. 22. Rennefahrt UE, et al. (2007) Specificity profiling of Pak kinases allows identification of 43. Zalc B, Goujet D, Colman D (2008) The origin of the myelination program in verte- novel phosphorylation sites. J Biol Chem 282:15667–15678. brates. Curr Biol 18:R511–512.

6of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1116560109 Strochlic et al. Downloaded by guest on September 26, 2021