Phosphorylation of the Bruchpilot N-terminus in Drosophila unlocks axonal transport of active zone building blocks

Rapid signal transduction within the nervous system underlies all cognitive processes, including memory formation and learning. Signals can be transmitted between neurons and from neurons to non-neuronal cells through chemical synapses: an arriving action potential triggers neurotransmitter release from synaptic vesicles (SVs) at specialized sites of a presynaptic compartment, called active zones (AZs), which are located in close proximity to a postsynaptic compartment that is equipped with neurotransmitter receptors (Südhof, 2012; Walter et al., 2018). Electron-dense membrane-associated protein scaffolds at the presynaptic AZs regulate the docking, priming, exocytic fusion and endocytic recovery of SVs. To this end, AZs can accumulate SVs, precisely define the sites of SV fusion and accurately position voltage-gated Ca²⁺ channels (Haucke et al., 2011).

Several hundred protein species populate the vicinity of the presynaptic AZs, with individual copy numbers varying over several orders of magnitude (Wilhelm et al., 2014). However, members of only a few protein families, including the CAST/ELKS, RIM, RIM-binding protein (RIM-BP, also known as RBP), Unc13 (also known as MUNC13 in mammals), Liprin-α and SYD-1 (also known as RhoGAP100F) families, comprise the presynaptic AZ scaffolds per se (Gundelfinger et al., 2015; Petzoldt and Sigrist, 2014; Südhof, 2012; Walter et al., 2014). AZ scaffold composition is evolutionarily conserved; for example, AZ scaffolds at Drosophila neuromuscular junctions (NMJs) form T-bar-shaped structures, also referred to as cytomatrices at the AZ (CAZs), which are organized around the central CAST/ELKS scaffolding protein Bruchpilot (BRP) and RIM-BP (Kittel et al., 2006; Liu et al., 2011; Van Vactor and Sigrist, 2017). Electron tomography and super-resolution light microscopy have helped to elucidate aspects of the molecular organization of presynaptic AZ scaffolds (Ackermann et al., 2015; Kittel et al., 2006; Kittelmann et al., 2013; Maglione and Sigrist, 2013).

New AZ scaffolds have to be assembled during the development of synaptic circuits and most likely are dynamic structures, in which components are turned over. AZ scaffold size and complexity in mammals and Drosophila are directly coupled to the resulting functional synaptic output (Ackermann et al., 2015). Modulation of the size and possibly the composition of AZ scaffolds might serve as a mechanism to adapt the SV release function to specific environmental demands over time scales ranging from minutes up to days (Van Vactor and Sigrist, 2017). Consistent with this notion, AZ scaffold assembly relies on dynamic protein–protein interactions based on conserved interaction surfaces, such as SAM, PDZ or coiled-coil domains (Südhof, 2012).

Presently, our knowledge of how scaffold components are transported to and assembled at newly forming AZs or how they might exchange with subunits of already established AZ scaffolds is fragmentary. Current studies, therefore, focus on clarifying the routes and kinetics involved in assembly and disassembly of AZ scaffolds. De novo assembly and turnover of AZ scaffolds presumably involve synthesis of AZ scaffold components in the cytoplasm of the neuronal cell body, and their transport over large distances along axons to synaptic terminals (Johnson et al., 2009; Nieratschker et al., 2009; Siebert et al., 2015). Electron and light microscopic analyses of AZ scaffolds provided some evidence for AZ scaffold formation from discrete building blocks (Ehmann et al., 2015; Kittelmann et al., 2013; Matkovic et al., 2013; Shapira et al., 2003). Axonal transport of pre-formed AZ building blocks thus provides a logical principle for AZ assembly mechanisms and dynamics (Bury and Sabo, 2016; Maeder et al., 2014). However, the unequivocal demonstration and compositional analysis of such transport packages is challenging, as they might exhibit low copy numbers of scaffold proteins. Moreover, it is presently unclear how premature aggregation of the scaffold components beyond the level of individual building blocks during transport would be prevented.

Previous work on the mechanisms underlying transport of AZ components through axons implicated serine-arginine protein kinase at 79D (SRPK79D) in this process. Knockout of SRPK79D or inactivation of its kinase activity led to the formation of interconnected, electron-dense, axonal aggregates that contained BRP and that resembled over-sized T-bars (Johnson et al., 2009; Nieratschker et al., 2009). Given that SRPK proteins are known to phosphorylate serine residues in serine-arginine dipeptide-rich regions (RS domains) of SR proteins, a family of regulatory factors involved in alternative pre-mRNA splicing and other gene regulatory processes (Lin and Fu, 2007; Zhou and Fu, 2013), the functional implication of SRPK79D in the transport of AZ scaffold components is surprising. Furthermore, as known AZ scaffold components do not contain canonical RS domains or extended RS dipeptide repeats, and as direct phosphorylation of these proteins by SRPK79D has not yet been demonstrated, it is presently unclear how SRPK79D might mechanistically intersect with the transport of AZ proteins.

Here, we delineated in vivo phosphorylation sites in BRP and found that BRP variants bearing a phosphorylation-defective N-terminus led to axonal T-bar-like assemblies that closely resembled aggregates elicited by SRPK79D mutants. The assemblies not only contained the major AZ components BRP and RIM-BP, but also comprised the critical release factor Unc13A (also known as UNC-13), suggesting co-transport of the three proteins as a pre-formed AZ scaffold building block. Systematic yeast two-hybrid (Y2H) screening, in vitro interaction studies and mass spectrometric analyses showed that SRPK79D specifically binds at the N-terminus of BRP and phosphorylates specific sites within this predicted unstructured region, including sites that upon mutation led to axonal aggregation. The homologous region of mammalian CAST/ELKS proteins was also phosphorylated by members of the SRPK family. Genetic analyses showed that phosphorylation of the N-terminal BRP stretch maintains transporting AZ building blocks in solution. Reversible phosphorylation of N-terminal regions in AZ scaffold proteins of the CAST/ELKS family by SRPKs might, therefore, constitute an evolutionarily conserved mechanism to stabilize the transport of a major AZ building block.

We hypothesized that reversible phosphorylation of AZ core components may prevent their precocious oligomerization, which poses a potential problem during their axonal transport. An N-terminal region of ∼300 residues in BRP appears to be largely intrinsically unstructured, whereas the remaining portions of BRP are predicted to adopt α-helical conformations, giving rise to an extended coiled-coil structure of the protein that shapes the T-bar AZ scaffold in its ultrastructural extensions (Fig. 1A) (Fouquet et al., 2009). Bioinformatics analyses using the NetPhos 3.1 server (Blom et al., 1999) predicted many phosphorylation sites clustered in the presumably unstructured N-termini of BRP and CAST/ELKS family AZ proteins (Fig. 1A,B).

To experimentally determine sites at which BRP is phosphorylated in vivo, synaptosomes were isolated from adult Drosophila fly head protein extracts as previously described by us (Depner et al., 2014). Monoclonal antibody NC82 (Fig. 1A) was used to immunoprecipitate BRP. The precipitate was fractionated by 12.5% SDS-PAGE, the prominent 190 kDa band was excised and subjected to tandem mass spectrometry (MS)-based peptide sequencing. This analysis revealed 27 phosphorylation sites in this long BRP-190 isoform (Fig. 1A). Notably, 13 of these phosphorylation sites fell within the N-terminal 140 residues of BRP-190, preceding the extended coiled-coil regions (Fig. 1A,B). The phosphorylation sites found within the coiled-coil regions predominantly reside in stretches that interrupt the predicted coiled-coil structures (Fig. 1A). Interestingly, several phosphorylation sites detected in Drosophila BRP are conserved and have also been shown to be phosphorylated in the mammalian BRP homologues CAST1 (also known as ERC2) and CAST2 (also known as ERC1 or ELKS) (Fig. 1A,B) (Dephoure et al., 2008; Hornbeck et al., 2015; Parker et al., 2015; Sacco et al., 2016; Sharma et al., 2014).

To delineate functional roles of BRP phosphorylation, we generated large, genomic p[acman] brp constructs, in which identified phosphorylation sites were abrogated by alanine mutations (Venken et al., 2006). We have previously shown that wild-type p[acman] brp constructs (brp^rescue) completely rescued brp^null (brp^Δ6.1/brp^DF(2R)69) alleles (per se pupal lethals) to full viability and fertility and are thus equivalent to the endogenous locus (Matkovic et al., 2013). We started by simultaneously exchanging S16, S32, S43, T59, Y130, S137, S629 and S1216 of BRP-190 for non-phosphorylatable alanines (brp^multiAla). The majority of these residues are not strongly conserved in the CAST/ELKS family (Fig. 1A,B). Similar to brp^rescue, expression of brp^multiAla elicited a complete rescue of brp^null alleles, indicating no essential function for phosphorylation at the altered sites.

Next, we generated a genomic p[acman] brp construct, in which S71, S73 and S90, which are highly conserved throughout the CAST/ELKS family (Mochida et al., 2016; Parker et al., 2015), were jointly exchanged for alanines, and introduced it into a brp^null background in comparison to a non-mutated control construct (Matkovic et al., 2013). Determination of hatching rates revealed robust differences between the control and the phosphorylation-defective brp^{SSS71/73/90AAA} mutant. The Mendelian ratio within the F1 generation of brp^{SSS71/73/90AAA} flies was about 13% (first replicate cross: n=105 animals, 13 brp^{SSS71/73/90AAA} flies, 92 wild-type flies; second replicate cross: n=178 animals, 23 brp^{SSS71/73/90AAA} flies, 155 wild-type flies), as compared to about 27% (first replicate cross: n=107 animals, 29 brp^rescue flies, 78 wild-type flies; second replicate cross: n=182 animals, 50 brp^rescue flies, 132 wild-type flies) for the non-mutated control (formally expected rate 33%). These observations suggest that the brp^{SSS71/73/90AAA} mutant is functionally compromised as compared to the rescue control.

Strikingly, confocal microscopy (Fig. 2A–C) revealed numerous BRP-positive (0.13 spots per µm² individual axon area) and RIM-BP-positive (0.15 spots per µm² individual axon area; Fig. 2D) aggregates within motoneuron axons of third instar larvae upon disruption of BRP N-terminal phosphorylation sites at positions 71, 73 and 90, suggesting disturbed axonal transport of AZ scaffold proteins. Most aggregates were positive for both BRP and RIM-BP. In contrast, brp^rescue (control) axons exhibited significantly fewer BRP-positive (0.005 spots per µm² individual axon area) and RIM-BP-positive (0.04 spots per µm² individual axon area; Fig. 2D) punctae. Furthermore, in brp^{SSS71/73/90AAA} mutant animals, the BRP (0.14 µm²) and RIM-BP (0.12 µm²) spots had a similar size and were significantly larger and abnormally formed compared to those found in control nerves (BRP 0.05 µm²; RIM-BP 0.06 µm²; Fig. 2E).

The brp^{SSS71/73/90AAA}-dependent axonal aggregates were reminiscent of previously reported BRP-positive and RIM-BP-positive axonal super T-bar structures that emerged upon loss of function of SRPK79D (Fig. 2C) (Nieratschker et al., 2009; Siebert et al., 2015). In srpk79D^VN mutants (srpk79D^VN bears a large deletion in the srpk79D gene, generating a srpk79D null mutant) (Johnson et al., 2009; Nieratschker et al., 2009), we found the average sizes of BRP-positive (0.3 µm²) and RIM-BP-positive (0.25 µm²) punctae further enlarged compared to corresponding punctae in brp^{SSS71/73/90AAA} mutants (Fig. 2E). In electron microscopy analyses, electron-dense aggregates were readily found in motoneuron axons of brp^{SSS71/73/90AAA} mutants, whose appearance and distribution closely resembled the aggregates forming in srpk79D^VN mutants (Fig. 3A,B). Likewise, stimulated emission depletion (STED) light microscopy at a resolution of ∼50 nm (Hell, 2007) showed a ‘stoichiometric patterning’ of BRP and RIM-BP epitopes within the brp^{SSS71/73/90AAA} aggregates, very similar to the previously reported, STED-visualized patterning of srpk79D mutant axonal aggregates (Fig. 3C,D) (Siebert et al., 2015).

Recent work by our group has shown that at AZs, the BRP–RIM-BP scaffold is crucial for proper clustering and positioning of the critical release factor Unc13A (Böhme et al., 2016). As for RIM-BP, staining for Unc13A revealed close and stoichiometric association of BRP and Unc13A in the axonal aggregates of both brp^{SSS71/73/90AAA} and srpk79D^VN mutants (Fig. 3E,F). In fact, the STED-resolved relative distribution of RIM-BP versus the C-terminus of BRP, as well as of Unc13A versus the C-terminus of BRP, were reminiscent of the organization of these epitopes within the scaffolds of synaptic AZs (Böhme et al., 2016; Liu et al., 2011).

The above morphological and compositional analyses strongly suggest that the axonal T-bar-like aggregates forming in srpk79D-deficient and brp^{SSS71/73/90AAA} mutants are equivalent in nature. To further characterize which AZ protein(s) cause these axonal T-bar-like aggregates, we genetically generated double mutant situations in which either rim-bp and srpk79D or brp and srpk79D were deleted. The brp locus gives rise to at least two prominent BRP isoforms, originating from alternative promoters, which contain (BRP-190) or lack (BRP-170) an N-terminal stretch of ∼320 amino acid residues (Fig. 1A). Previous analyses have shown that AZ scaffolds comprise a circular arrangement of alternating BRP-190 and BRP-170 clusters (Matkovic et al., 2013). To test whether both isoforms are needed for the formation of the axonal T-bar-like aggregates, isoform-specific brp mutants (Matkovic et al., 2013) were tested in a srpk79D^atc background (srpk79D^atc leads to the production of a non-functional, truncated form of SRPK79D; Johnson et al., 2009). No axonal aggregates were formed when both BRP isoforms or BRP-190 alone were knocked out in a brp^null background (Fig. 4A–E). In contrast, axonal aggregates were still observed when only BRP-170 was knocked out (Fig. 4F,G). Likewise, rim-bp and srpk79D double knockouts still displayed BRP-positive aggregates, indicating that RIM-BP is not essential for axonal aggregate formation (Fig. 4H–J). These results suggest that BRP-190 might be the only AZ scaffold component that is essential for the formation of axonal T-bar-like aggregates upon knockout or mutation of SRPK79D or, most likely, in brp^{SSS71/73/90AAA} mutants. Thus, the N-terminal ∼320 residues that discriminate BRP-190 from BRP-170 seem to play an important and specific role in the formation of BRP-positive axonal aggregates. Our observations, thus, suggest a functional relationship between SRPK79D and the N-terminal region of BRP-190 in preventing axonal aggregation of AZ components during their transport to synapses.

As the brp^{SSS71/73/90AAA} mutant phenocopies the srpk79D^VN mutant, axonal aggregates seen in these mutants likely originate as a result of the same mechanistic principle. The simplest explanation for the similar effects of these mutants would be that SRPK79D phosphorylates BRP to facilitate axonal transport, although BRP does not contain a typical RS domain. This hypothesis is supported by the observation that SRPK79D and BRP co-localize in vivo (Johnson et al., 2009). We directly tested this idea by conducting in vitro binding and phosphorylation studies using recombinant SRPK79D and BRP fragments.

While SRPKs encompass the canonical N- and C-terminal lobes of Ser-Thr protein kinases, a region of ∼200 residues that is predicted to be intrinsically disordered intervenes between the lobes in this protein kinase family (Ghosh and Adams, 2011). In addition, a shorter region of predicted intrinsic disorder precedes the N-terminal lobe of these kinases. Moreover, some SRPKs have been shown to engage their substrates via a docking groove on the C-lobe (Ngo et al., 2005). SRPK79D contains a split kinase domain typical of the SRPK family and appears to encompass a docking groove, but harbors a significantly longer region of predicted intrinsic disorder preceding the N-terminal lobe compared to mammalian SRPK1 and SRPK2 (Fig. 5A). The SRPK79D unstructured N-terminus, which is required for its localization with BRP in vivo (Johnson et al., 2009), and its putative docking groove on the C-lobe represent possible regions through which the enzyme might transiently engage substrate proteins. Engaged regions can be distinct from or overlapping with the phosphorylated regions. In addition, SRPK79D might also use these regions to dock to other proteins that in turn are associated with substrate proteins. We therefore employed the Y2H system (Worseck et al., 2012) to uncover potential direct interactions between SRPK79D and the proteins found in the axonal aggregates. In addition to full-length proteins, we tested Y2H interactions among fragments of the proteins that covered known functional regions, predicted or known folded domains or regions of predicted intrinsic disorder.

A robust Y2H interaction was detected between full-length SRPK79D (SRPK79D^FL) as well as all SRPK79D constructs containing the two lobes of the kinase core (SRPK79D^Core, SRPK79D^{CoreΔlinker1}, SRPK79D^{CoreΔlinker2}) and a region comprising the N-terminal 152 residues of the BRP-190 isoform (BRP-190^1–152; Table 1). Neither the intrinsically disordered SRPK79D N-terminal region (SRPK79D^1–340) nor the linker region between the N- and C-lobes (omitted in constructs SRPK79D^{CoreΔlinker1} and SRPK79D^{CoreΔlinker2}) were required for Y2H interactions with BRP-190^1–152. The SRPK79D N-terminus alone interacted with diverse, putatively unstructured or coiled-coil regions of tested AZ proteins, among these BRP and RIM-BP. These results indicate that BRP-190 might directly interact with SRPK79D, possibly via its N-terminal region inserting into the docking groove of the kinase.

To test whether the observed Y2H interactions of SRPK79D and BRP-190^1–152 originated from direct contacts between these proteins, we produced recombinant protein fragments (Fig. 5A), and tested their interactions by analytical size-exclusion chromatography (SEC). To test for the role of the docking groove in the interaction, we generated a SRPK79D variant bearing a disrupted docking groove (SRPK79D^CoreΔdock) (Lukasiewicz et al., 2007). SRPK79D^Core and BRP-190^1–152 co-migrated during SEC and together eluted earlier than the isolated proteins (Fig. 5B). Consistent with the Y2H results, truncation of the inter-lobe linker region of SRPK79D (SRPK79D^{CoreΔlinker1}) had no effect on BRP-190^1–152 binding (data not shown). In line with the idea that BRP-190^1–152 binds to the docking groove of SRPK79D, the SRPK79D^CoreΔdock variant did not co-elute with BRP-190^1–152 in SEC (Fig. 5C).

To further narrow down the SRPK79D binding site(s) on BRP-190^1–152, we conducted peptide SPOT analyses. SPOT membranes contained overlapping 25-residue peptides covering the BRP-190 N-terminal region with a seven-residue offset (Table S2). The SRPK79D kinase domain construct (SRPK79D^Core) bound specifically to peptides representing two BRP regions (66-HHRSRSASR-74; 113-RSRDRSLER-121; Fig. 5E), suggesting that it can attach to several sites on BRP-190. Notably, the sequences of these two putative docking sites resemble proposed binding motifs recognized by SRPK1, which consist of three basic residues (bold) separated by two to three residues (Lukasiewicz et al., 2007). Based on a similar sequence fingerprint, we tentatively assigned a possible third binding motif for SRPK79D in the BRP-190 N-terminus (15-RSPGRVRR-22), although this region also showed binding to the GST control (Fig. 5E). Taken together, these data suggest that direct binding of SRPK79D within the ∼150 N-terminal residues of BRP-190 might underlie the functional SRPK79D–BRP interplay in preventing axonal T-bar-like aggregate formation.

SRPKs can employ different modes of operation depending on the nature of the substrate proteins. Substrates comprising extended RS repeats can be phosphorylated via a processive mechanism that involves an initial engagement of the substrate at a docking groove on the C-terminal lobe of the kinase (Lukasiewicz et al., 2007). Subsequently, phosphorylation sites are continuously funneled into the active center in a C-to-N-terminal direction. Continued phosphorylation eventually reduces target affinity, most likely by electrostatic repulsion between the phosphates and the acidic docking groove (Ghosh and Adams, 2011). For targets with only short RS repeats, docking groove binding can be dispensable and such targets can be phosphorylated via a distributive mechanism or in a dual-track mode that encompasses processive and distributive phases (Aubol et al., 2013; Lukasiewicz et al., 2007).

To test whether the observed interaction of SRPK79D with the N-terminal region of BRP-190 forms the basis for SRPK79D-mediated BRP phosphorylation, we conducted in vitro phosphorylation assays using recombinant proteins. Upon incubation with γ-[³²P]-ATP, SRPK79D^Core phosphorylated full-length BRP-190, BRP-190^1–152 as well as itself, but strikingly failed to phosphorylate a large BRP-190 fragment (BRP190^Δ1–152) that lacked the N-terminal 152 residues (Fig. 5F). Identical preparations of catalytically inactive SRPK79D^Core-dead, bearing a K376M exchange that disrupts ATP binding (Johnson et al., 2009), still bound BRP-1901-152 (Fig. 5D) but did not exhibit similar kinase activity (Fig. 5F), indicating that the observed phosphoryl-transfer activity originates from recombinant SRPK79D^Core. In contrast, the SRPK79D^CoreΔdock variant, which exhibits a disrupted docking groove, phosphorylated the same BRP variants as SRPK79D^Core (Fig. 5F). These results are consistent with the idea that SRPK79D specifically targets the BRP N-terminal region, and that the enzyme can act through a distributive mechanism, as previously described for SRPK1 (Aubol et al., 2013).

To determine the sites of SRPK79D^Core-mediated phosphorylation of BRP, we analyzed non-phosphorylated and in vitro phosphorylated BRP-190^1–152 by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS). After one hour of incubation with SRPK79D^Core or SRPK79D^CoreΔdock and ATP, BRP-190^1–152 showed an increase in molecular mass of 556 Da, suggesting seven added phosphate groups (each contributing ∼80 Da), with a smaller portion of the protein carrying eight phosphate groups (Fig. 6A). Time course experiments monitoring BRP-190^1–152 phosphorylation by SRPK79D^Core and SRPK79D^CoreΔdock revealed faster phosphorylation with an intact SRPK docking groove (Fig. S1). Independent of the docking groove, the same phosphorylation state was reached after one hour (Fig. S1, Fig. S2A). These findings suggest that an intact docking groove increases phosphorylation kinetics but does not alter phosphorylation specificity.

To map the exact phosphorylation sites in BRP-190^1–152, we performed tryptic in-gel digestion followed by mass spectrometry. By using a combination of liquid chromatography electrospray ionization mass spectrometry (LC-ESI-MS) and MALDI-TOF-MS, we confirmed almost complete phosphorylation of residues S16, S32, S34, S71, S73, S90 and S118 after one hour of incubation (Fig. 6B,C). Except for S32, S34 and S90, these residues reside within one of the putative SRPK79D docking sites on BRP-190^1–152. S90 is the only site that is not part of a RS or SR dipeptide.

Analyzing the time course of SRPK79D^Core-mediated BRP-190^1–152 phosphorylation in vitro by LC-ESI-MS showed that S90 and S118 are phosphorylated first; more than 90% of S90 residues and more than 80% of S118 residues were phosphorylated within 30 s. S71/S73, S32/S34 and S16 were phosphorylated at progressively lower rates (Fig. 6B). The exact order of phosphorylation of the closely spaced residues could not be resolved, as S71/S73 on the one hand and S34/S32 on the other were part of the same tryptic peptides.

Our MS analyses reliably covered over 90% of the BRP-190^1–152 sequence and almost all of its potential 7–8 phosphorylation sites observed by analysis of the intact BRP-190^1–152 using MALDI-TOF-MS (Fig. 6B,C). However, we did not observe peptides containing S114, which resides within one of the putative binding motifs in an RSR sequence, possibly because of several arginine residues in the immediate vicinity giving rise to very small tryptic peptides. To confirm phosphorylation sites that were not covered by detectable tryptic peptides, we expressed recombinant sub-fragments of BRP-190^1–152 and incubated them with SRPK79D^Core or SRPK79D^CoreΔdock in the presence of ATP. The phosphorylation state of these sub-fragments was then analyzed by MALDI-TOF-MS, indicating almost complete phosphorylation at S114 and partial phosphorylation at S16 (Fig. S2).

The above findings show that SRPK79D predominantly phosphorylates sites in BRP-190^1–152 that reside within, or in the direct vicinity of, the kinase docking sites. SRPK79D-mediated phosphorylation seems to preferentially start at the C-terminal region of BRP-190^1–152, around S90 and S118. The directional C-to-N-terminal phosphorylation, together with faster phosphorylation with an intact docking groove, hints at the possibility that SRPK79D might also be able to work processively for faster phosphorylation of several sites but can still work in a distributive manner (docking groove mutant), as previously seen for SRPK1 and SRSF1 (Aubol et al., 2003).

Notably, all the identified in vitro phosphorylation sites with the exception of S114 and S118 correspond to sites that were also found phosphorylated in vivo; in particular, they encompass S71, S73 and S90, which upon alanine mutation led to axonal BRP–RIM-BP–Unc13A aggregates (see above). None of the additional, potentially phosphorylatable seven serines and five threonines within the first 152 residues of BRP were phosphorylated significantly by SRPK79D in vitro, suggesting that the recombinant SRPK79D constructs largely retain the substrate specificity of endogenous SRPK79D. Seven in vivo sites at the BRP-190 N-terminus (S43, T59, S60, Y130, S133, S137 and S139) were not found to be phosphorylated by SRPK79D in vitro and may thus be targeted by other kinases (Fig. 6C).

As ongoing phosphorylation of target sites reduces the affinity of substrate proteins to SRPK1 (Ghosh and Adams, 2011), phosphorylated BRP-190 might likewise exhibit reduced affinity to SRPK79D. To test this prediction, we monitored the interaction of SRPK79D^Core with BRP-190^1–152 by analytical SEC after prior SRPK79D^Core-mediated phosphorylation of BRP-190^1–152. Phosphorylation of BRP-190^1–152 at positions S32, S34, S71, S73, S90 and S118 by SRPK79D^Core in the analyzed samples was verified by MALDI-TOF-MS after digestion with trypsin. In contrast to non-phosphorylated BRP-190^1–152, the phosphorylated BRP-190^1–152 no longer co-migrated with SRPK79D^Core in SEC (Fig. 5B). These results suggest that phosphorylation reduces affinity of the substrate for the SRPK79D docking groove. Reduced affinity of the substrate upon its phosphorylation additionally supports mechanistic commonalities between SRPK79D and SRPK1.

To test if mammalian SRPKs can phosphorylate the N-termini of mammalian BRP orthologues CAST1 and CAST2, we conducted in vitro binding studies and kinase assays, using recombinant SRPK1, SRPK2, CAST1 and CAST2 constructs. Two fragments comprising N-terminal regions of CAST2 (CAST2^1–163 and CAST2^1–353) co-migrated with SRPK1^CoreΔlinker (equivalent to SRPK1ΔNS1 in Ngo et al., 2005) and SRPK2^Core, eluting earlier from the column than the individual proteins (Fig. 6D,E). Moreover, both fragments, as well as the related BRP-190^1–152, were phosphorylated by the kinases upon incubation with γ-[³²P]-ATP (Fig. 6F). These data are consistent with the idea that SRPK-mediated phosphorylation of the N-termini of CAST/ELKS family proteins is an evolutionarily conserved regulatory principle that may be involved in controlling axonal transport of AZ scaffold components in neurons across the animal kingdom.

The molecular mechanisms underlying ordered axonal transport of AZ scaffold proteins are presently poorly understood. In particular, it is unclear whether, and if so which, AZ scaffold components are co-transported as functional building blocks, and how neurons avoid pre-mature higher-order aggregation of AZ scaffold proteins during transport. Several lines of evidence indirectly implicate reversible protein phosphorylation as an important regulatory principle for axonal transport of AZ scaffold proteins. Interestingly, several kinase families, such as protein kinase A, Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) and glycogen synthase kinase 3 (GSK3), seem to negatively regulate transport of AZ and SV material, as their knockout has been shown to promote transport (Guillaud et al., 2008; Hall and Hedgecock, 1991; Morfini et al., 2002; Sato-Yoshitake et al., 1992; Wairkar et al., 2009). In contrast, functional impairment of SRPK79D in Drosophila causes ectopic accumulation of scaffold proteins within the axoplasm (Johnson et al., 2009; Nieratschker et al., 2009), indicating that SRPK79D is a positive regulator of axonal transport. However, understanding of how SRPK79D supports transport of AZ scaffold proteins has so far remained elusive.

Here, we show that SRPK79D mediates a ‘transport-stabilizing’ function by phosphorylating a specific ∼150-residue stretch at the N-terminus of the extended coiled-coil domain protein BRP. SRPK79D can bind at arginine-rich motifs in this region of BRP and phosphorylates at least seven sites within or in close vicinity to these motifs. When interfering with SRPK79D-mediated phosphorylation of BRP by exchanges from serine to non-phosphorylatable alanine in the brp genomic context, axonal aggregates formed. The aggregates resembled those that form on knockdown of SPRK79D in srpk79D^VN mutants in ultrastructural detail and molecular composition: large, extended, multiple T-bars containing BRP, RIM-BP and Unc13A. Importantly, these aggregates no longer formed when the BRP-190 isoform exclusively containing this N-terminal sequence stretch was genetically eliminated. The sizes of the BRP–RIM-BP–Unc13A aggregates in the brp^{SSS71,73,90AAA} mutant were somewhat reduced compared to the qualitatively very similar aggregates in sprk79D mutants. Thus, it is possible that the sites not included in our triple mutant but found to be phosphorylated by SRPK79D in vitro (S16, S32, S34, S114, S118) also contribute to the srpk79D mutant phenotype. Furthermore, it is likely that kinases other than SRPK79D contribute to phosphorylation of BRP-190, as suggested by additional phosphorylation sites identified in vivo (S43, T59, S60, Y130, S133, S137, S139). Nevertheless, it is clearly a key function of SRPK79D to keep the N-terminal stretch of BRP-190 phosphorylated during axonal transport. It thereby protects the transported BRP-190 isoform from undergoing a pseudo-AZ-like assembly process within the axoplasm, which co-aggregates, but does not depend on, RIM-BP and Unc13A. The BRP-170 isoform lacks the N-terminal region that is reversibly phosphorylated in BRP-190 and, thus, cannot be phosphorylated by SRPK79D. It appears likely that the transport of the two isoforms is uncoupled, at least to a large degree.

Further expanding on our previous analyses (Siebert et al., 2015), results reported here document that brp^{SSS71,73,90AAA}- and srpk79D^null-induced aggregates accumulate only a specific sub-spectrum of AZ proteins, namely BRP, RIM-BP and Unc13A. These three proteins exactly constitute SV release sites within the central AZ scaffold (Reddy-Alla et al., 2017). They incorporate into assembling AZ scaffolds only after other scaffold proteins, SYD-1 and Liprin-α, have initialized the actual assembly process (Böhme et al., 2016). We did not find the ‘early seeding factors’ SYD-1 and Liprin-α within the aggregates observed in either the srpk79D or the brp^{SSS71,73,90AAA} mutants. Thus, BRP phosphorylation by SRPK79D seems to block the premature oligomerization of a specific scaffold building block (BRP–RIM-BP–Unc13A). We previously showed that RIM-BP is directly transported via a high-affinity interaction between its SH3 domains II and III and a proline-rich stretch of the JIP-1 homologue Aplip1, which in turn binds to the kinesin 1A-type motor Unc-104 (Siebert et al., 2015). Thus, the RIM-BP constituent might actually provide a key connection between the transported building block and the transport machinery.

To the best of our knowledge, our work documents for the first time direct phosphorylation and consequent regulation of a synaptic protein by a SRPK-type kinase. SRPK family members have so far been predominantly implicated in the phosphorylation, and thus regulation, of SR proteins that act as regulators of various gene expression processes (Lin et al., 2007; Zhou and Fu, 2013). Our mechanistic results imply a similar mode of action of SRPK79D on BRP as previously described for SRPKs acting on RS domain-containing substrates. Although the docking groove seems to be dispensable for the overall phosphorylation state of the BRP N-terminus, our data show faster phosphorylation with an SRPK79D variant bearing an intact docking groove. SRPK79D with an intact docking groove also shows a tendency to phosphorylate the more C-terminal region of BRP-190^1–152 first, indicated by a faster phosphorylation rate of S90 and S118 compared to the more N-terminally located sites.

Consistent with our results, it has been shown that the non-conserved, putatively unstructured N-terminus of SRPK79D is important for its localization with BRP (Johnson et al., 2009). Our Y2H studies further show that this SRPK79D N-terminal portion engages in several weak and rather unspecific interactions with AZ proteins in predicted unstructured or coiled-coil regions, notably also in the co-transported proteins BPR and RIM-BP. We suggest that these weaker interactions do not directly impact on the SRPK79D mechanism of action, but might rather ensure a high local concentration of SRPK79D at the transported protein complex.

The question arises, how might the phosphorylation status of a confined, likely intrinsically disordered stretch at the BRP N-terminus influence the aggregation status of an entire transport package that, apart from BRP, contains RIM-BP and Unc13A? We speculate that the high charge density at the BRP N-terminus introduced through multiple phosphate moieties might trigger an extensive, cooperative conformational switch in the protein, rendering it less prone to formation of aggregations. Consistent with this notion, conformational changes can be propagated over long distances through coiled-coil arrangements in proteins, as illustrated, for example, by ATP-binding–hydrolysis–ADP-P_i release cycles in motor proteins (Carter et al., 2016). As an alternative mechanism, EEA1 has recently been shown to undergo a massive extended-to-collapsed conformational change that is initiated upon binding to Rab5:GTP and that is propagated over the entire length of this 1400-residue protein (Murray et al., 2016).

Ultimately, the BRP–RIM-BP–Unc13A building block transported down the axon must be integrated into the AZ scaffold. It appears likely that local dephosphorylation of BRP might be part of the integration process. Thus, one might expect localized phosphatase activity to promote scaffold assembly. In fact, in previous synaptic AZ assembly studies at Drosophila NMJ synapses, protein phosphatase 2A (PP2A) complex proteins were found to regulate presynaptic assembly. In the absence of the phosphatase, assembling postsynaptic glutamate receptor fields often lacked presynaptic AZ scaffolds (Viquez et al., 2009), a finding that is at least consistent with PP2A supporting developmental scaffold assembly via BRP dephosphorylation. PP2A activity and assembly function are seemingly tuned by activities of the serine-threonine kinase GSK3B (Viquez et al., 2009) and Unc-51 (Atg1) (Wairkar et al., 2009).

We also show that mammalian BRP homologues are equally phosphorylated by SRPKs at their conserved N-terminal stretches. Notably, SRPKs have been implicated in various neurodegenerative diseases (Chan and Ye, 2013; Jang et al., 2009). Furthermore, mammalian SRPK2 has already been shown to play a role in neuronal function by phosphorylating the tau protein (also known as MAPT) at a specific position to inhibit axonal elongation in neurons (Hong et al., 2012). Our results, therefore, might be of importance for AZ assembly and plasticity and, consequently, developmental circuit formation, and learning and memory processes in the human brain. In addition, they further underscore that future studies investigating the molecular principles underlying SRPK-related neuronal diseases (Chan and Ye, 2013) should take into account not only well-documented functions of SRPKs in regulating gene expression but also roles by which these enzymes might more directly influence the functions of neuronal proteins.

DNA fragments encoding SRPK79D constructs from a codon-optimized gene (Centic Biotec) were cloned into pGEX 6P1 vector using BamHI and NotI and altered by QuikChange mutagenesis (Agilent) where necessary. A DNA fragment encoding SRPK2^Core from a mSRPK2 plasmid was provided by Susanne Schoch-McGovern (Universität Bonn, Germany) was cloned into pETM11 vector (EMBL, Heidelberg), using NcoI and SalI. SRPK1 expression constructs were provided by Gourisankar Ghosh (University of California, San Diego, USA). DNA fragments encoding BRP N-terminal constructs from a codon-optimized gene (Centic Biotec) were cloned into pETM11 vector using NcoI and SalI and altered by QuikChange mutagenesis where necessary. DNA fragments encoding BRP-190 and BRP-190^Δ1–152 were cloned from a codon-optimized gene (Centic Biotec) into a modified pFL vector (EMBL, Grenoble) that directed production of protein bearing an N-terminal His₁₀-tag followed by a TEV cleavage site and a C-terminal Strep-tag, using EcoRI and SalI. A DNA fragment encoding CAST1 was cloned from cDNA into a modified pFL vector that directed production of protein bearing an N-terminal His₁₀-tag followed by a TEV cleavage site, using EcoRI and SalI. DNA fragments encoding CAST2 N-terminal fragments were cloned from rat cDNA into pETM11 vector, using NcoI and SalI. A DNA fragment encoding BRP-190^1–152,6SD was obtained as a synthetic gene (GeneArt, ThermoFisher) in a pET151/D-TOPO expression vector. All constructs were verified by DNA sequencing.

Protein constructs used here are listed in Table S1. Production of SRPK79D constructs was done in Escherichia coli BL21 Rosetta 2 cells in ZYM auto-induction media (Studier, 2005). Cells were grown for 4 h at 37°C and subsequently incubated at 18°C overnight. Cells were harvested by centrifugation (9000 g, 7 min) and resuspended in lysis buffer (400 mM NaCl, 40 mM Tris-HCl pH 7.5, 5% v/v glycerol, 1 mM DTT) supplemented with DNase. The cells were lysed by sonication (Sonopuls HD 3100, Bandelin) and lysate was cleared by centrifugation (55,000 g, 1 h, 4°C). Cleared lysate was incubated for 1 h with glutathione-sepharose 4B resin (GE Healthcare), washed with lysis buffer and protein was eluted in steps by adding 10 mM reduced glutathione to the lysis buffer. Tags were cleaved by adding 1:20 PreScission protease (GE Healthcare) and dialyzing against SEC buffer (200 mM NaCl, 20 mM Tris-HCl pH 7.5, 1 mM DTT) overnight. Depending on the size of the protein construct, the GST-tag was removed by recycling over glutathione–sepharose or in a final SEC step (Superdex 75 16/60, GE Healthcare). Proteins were concentrated and flash frozen in liquid nitrogen at concentrations of 10–20 mg/ml.

Constructs of the BRP N-terminus and the CAST2 N-terminus were produced and cells were lysed as described above, with addition of 20 mM imidazole to the lysis buffer. Affinity chromatography was carried out using Ni²⁺-NTA agarose (Macherey Nagel). Prior to protein loading, the resin was equilibrated with lysis buffer and the protein was eluted by addition of 300 mM imidazole to the lysis buffer. His tags were cleaved by adding 1:20 TEV protease overnight. Further purification was performed by ion-exchange chromatography on MonoS and MonoQ columns (GE Healthcare). Proteins were eluted by applying a linear salt gradient to 200 mM NaCl. Purified proteins were concentrated to 1–2 mg/ml and flash frozen in liquid nitrogen.

SRPK1 and SRPK2 constructs were expressed and purified by Ni²⁺-NTA affinity chromatography as described above. His tags were not cleaved, and final purification was done by SEC on Superdex S200 16/60 or 10/300 increase columns (GE Healthcare). Purified proteins were concentrated to 5–20 mg/ml and flash frozen in liquid nitrogen.

For production of BRP and CAST1 variants via recombinant baculoviruses in insect cells, E. coli DH10MultiBac cells were used to generate bacmids. SF9 cells were transfected with the purified bacmids for each construct and a first virus generation (V₀) was harvested after 72 h. V₀ virus was used to generate a virus with a higher titer (V₁) in SF9 cells, which was then used for large-scale production in High Five cells. Cells were harvested when viability dropped below 90% or when the eYFP signal reached a plateau. Harvested cells were either flash frozen in liquid nitrogen and stored at −80°C or directly used for purification.

Cell pellets of BRP-190 and BRP-190^Δ1–152 expression were resuspended in 400 mM NaCl, 40 mM Tris-HCl pH 8.5, 5% (v/v) glycerol, 1 mM DTT supplemented with protease inhibitors (Roche), and lysed by sonication. Lysates were cleared by centrifugation and proteins were captured on Strep-Tactin resin (IBA). After washing, proteins were eluted by addition of 2.5 mM desthiobiotin in 400 mM NaCl, 40 mM Tris-HCl pH 8.5, 5% (v/v) glycerol, 1 mM DTT and flash frozen in liquid nitrogen.

CAST1 was purified via Ni²⁺-NTA affinity chromatography, MonoQ ion exchange chromatography (elution in a linear salt gradient to 500 mM NaCl). Final SEC was carried out on a Superdex S200 10/300 column in 200 mM NaCl, 20 mM Tris-HCl pH 7.5, 1 mM DTT. Purified protein was concentrated to 0.8 mg/ml and flash frozen in liquid nitrogen.

Fly strains were reared under standard laboratory conditions (Sigrist et al., 2003) at 25°C, 65%–70% humidity and constant 12/12 h light/dark cycle in incubators. Both male and female larvae were used for analysis in all experiments. The following genotypes were used: (1) +/+ (w1118, WT), (2) srpk79D^VN/srpk79D^VN (srpk79D^VN), (3) srpk79D^atc/srpk79D^atc (srpk79D^atc), (4) Df(2R)BSC29/+; srpk79D^atc/srpk79D^atc (brp^Df/+; srpk79D^atc), (5) brp⁶⁹/Df(2R)BSC29; srpk79D^atc/srpk79D^atc (brp^null/brp^Df; srpk79D^atc), (6) Df S2.01/+; srpk79D^atc/srpk79D^atc (rim-bp^Df/+; srpk79D^atc), (7) rim-bp^STOP1/Df S201; srpk79D^atc/srpk79D^atc (rim-bp^null/rim-bp^Df; srpk79D^atc). Genomic brp^p[acman]WT was crossed to brp^{null Δ6.1}/brp^Df69 and genomic phosphorylation mutant brp^{SSS71/73/90AAA} was crossed to brp^{null Δ6.1}/brp^Df69. Stocks were obtained from: brp⁶⁹ (Kittel et al., 2006); Df2.01 and rim-bp^STOP1 (Liu et al., 2011); srpk79D^atc; srpk79D^vn (Nieratschker et al., 2009); genomic brp^p[acman]WT (Matkovic et al., 2013).

The attB-p[acman] BAC clone containing the genomic region of BRP was obtained from Matkovic et al. (2013). Mutations were incorporated according to the Counter Selection BAC Modification Kit (Gene Bridges GmbH) by using the following primers: Amplification of the rpsL cassette: 5′-CGACATGGATGAGCCAACCAGTCCGGCCGGAGCGGGTCACCATCGCAGCCGGGGCCTGGTGATGATGGCGGG-3′ (forward); 5′-GAATGGGTATGAACTCGCGATCATGGGGATCCACGAGTCCACCGCGATCCAGTCAGAAGAACTCGTCAAGAAG-3′ (reverse); mutagenesis 5′-ACATGGATGAGCCAACCAGTCCGGCCGGAGGGGTCACCATCGCAGCCGGGCCGCCGCCAGACCACCGATGGCCCATGCC-3′ (forward); 5′-ATGGGTATGAACTCGCGATCATGGGGATCCACGAGTCCACCGCGATCCAGCGCTTGGTAGCGGGTTC-3′ (reverse). After sequencing, the construct was injected into an attP site-containing fly strain (y[1] w[1118]; PBac[y[+]-attP-9A]VK00005; Bloomington Drosophila Stock Center line #9725) using the services of BestGene Inc.

For the identification of BRP residues phosphorylated in vivo, proteins were extracted from wild-type Drosophila heads in the presence of phosphatase inhibitors (PhosStop, Roche), immunoprecipitated with rabbit anti-BRP^last200 antibodies (20 μg coupled to 30 μl of protein A bead slurry [Bio-Rad]; Ullrich et al., 2015) and subjected to MS-based analyses, as previously described (Owald et al., 2010). Briefly, synaptosome membranes were enriched by differential centrifugation (Depner et al., 2014). Antibody-coupled resin (anti-BRPlast200 or identically coupled IgG control [catalog no. M8642-1MG, Sigma-Aldrich]) was incubated with solubilized and precleared synaptosome membrane preparations (LP1; 1 mg protein at 2 μg/μl) in IP buffer (20 mM HEPES pH 7.4, 200 mM KCl, 2 mM MgCl2, 1% Triton X-100) for 10 h at 4°C. Antibody-coupled resin was incubated with solubilized and precleared synaptosome membrane preparations (LP1; 1 mg protein at 2 µg/µl) in IP buffer (20 mM HEPES pH 7.4, 200 mM KCl, 2 mM MgCl₂, 1% Triton X-100) for 10 h at 4°C. After washing four times with IP buffer, Ab–Ag complexes were eluted with 60 µl 2× sample buffer (containing 1 M Tris-HCl pH 6,8; 10% SDS; glycerol; β-mercaptoethanol; 1% Bromphenol Blue). Samples were analyzed by western blot and subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis as described below. MS data were searched against the FlyBase (http://flybase.org/) database using the MASCOT search algorithm.

Yeast two-hybrid (Y2H) analyses were performed as described in Böhme et al. (2016). Briefly, DNA fragments encoding various regions of SRPK79D, BRP and RIM-BP were each cloned into two bait and two prey vectors. Diploid yeasts, carrying each a unique bait–prey vector pair, were generated by mating yeast carrying individual bait and prey vectors. Putative protein–protein interactions (PPIs) were identified by yeast growth on selective media (Worseck et al., 2012). Bait and prey constructs that led to auto-activation were removed from the analysis. Only bait vector–prey vector combinations that showed growth at least in four independent replicas were considered as putative PPIs. Y2H interactions are listed in Table 1.

Analytical size-exclusion chromatography (SEC) was performed on an 3.2/300 Superdex 200 increase column (GE Healthcare) at 4°C in 200 mM NaCl, 20 mM Tris-HCl pH 7.5, 2 mM MgCl₂, 1 mM DTT. 50–100 µg of SRPK variant was mixed with a 1.2-fold molar excess of BRP or CAST/ELKS family constructs. For complex formation, samples were incubated for 30 min on ice before loading onto the column. For runs involving prior phosphorylation, 1 mM ATP was added to the protein mixtures and samples were incubated for 30 min on ice. Fractions were collected and analyzed by 12.5% SDS-PAGE and visualized by Coomassie Brilliant Blue staining.

25 pmol of SRPKs were mixed with 25 pmol of a phosphorylation target in 20 µl reaction buffer (200 mM NaCl, 20 mM HEPES-NaOH pH 7.5, 2 mM MgCl₂, 1 mM DTT). Reactions were started by addition of 2 µl of a 2 mM ATP solution in reaction buffer supplemented with γ-[³²P]-ATP (9.25 MBq, 250 µCi). Samples were incubated for 1 h at room temperature and reactions were stopped by addition of SDS sample buffer and subsequent heating of the samples to 95°C for 5 min. Samples were analyzed by 12.5% SDS-PAGE and visualized by Coomassie Brilliant Blue staining and by scanning on a Storm PhosphorImager (GE Healthcare).

Time-course experiments were performed to analyze the SRPK97D-dependent phosphorylation of the BRP N-terminus. 3 µg (180 pmol) of BRP-190^1–152 were incubated in vitro with equal amounts of SRPK79D^Core or SRPK79D^CoreΔdock in 200 mM NaCl, 40 mM HEPES pH 7.5, 2 mM MgCl₂ on ice. The reactions were started by addition of 1 mM ATP. Samples were collected at time points of 30 s, 5 min and 1 h and stopped by addition of SDS sample buffer and boiling for 10 min at 95°C.

Samples were separated by 12.5% SDS-PAGE and proteins were stained by Coomassie Brilliant Blue. Gel bands corresponding to BRP-190^1–152 were excised, washed with 50% (v/v) acetonitrile in 50 mM ammonium bicarbonate, shrunk by dehydration in acetonitrile and dried in a vacuum centrifuge. The dried gel pieces were incubated with 50 ng trypsin (sequencing grade modified, Promega) in 25 µl of 50 mM ammonium bicarbonate at 37°C overnight. To extract the peptides, 25 µl of 0.5% (v/v) trifluoroacetic acid (TFA) in acetonitrile was added and the extract was dried under vacuum.

Peptides were transferred to 10 μl of 0.1% (v/v) TFA, 5% (v/v) acetonitrile and 2 µl were analyzed on a reversed-phase capillary nano liquid chromatography system (Ultimate 3000, Thermo Scientific) connected to an Orbitrap Velos mass spectrometer (Thermo Scientific). Samples were desalted on a trap column (PepMap100 C18, 3 μm, 100 Å, 75 μm i.d.×2 cm; Thermo Scientific) using a mobile phase of 0.05% TFA, 2% acetonitrile in water. After switching the trap column inline, LC separations were performed on a capillary column (Acclaim PepMap100 C18, 2 μm, 100 Å, 75 μm i.d.×25 cm, Thermo Scientific) at an eluent flow rate of 300 nl/min. Mobile phase A contained 0.1% formic acid in water, mobile phase B contained 0.1% formic acid in acetonitrile. The column was pre-equilibrated with 3% mobile phase B followed by an increase to 50% mobile phase B in 50 min. Mass spectra were acquired in a data-dependent mode, utilizing a single MS survey scan (m/z 350–1500) with a resolution of 60,000 in the Orbitrap, and MS/MS scans of the 20 most intense precursor ions in the linear trap quadrupole.

Identification of proteins was performed using the Mascot Daemon and Mascot Server version 2.5.0 (Matrix Science). Raw data were searched against an in-house custom protein sequence database including the sequence of BRP-190^1–152. A maximum of two missed cleavages was allowed and the mass tolerance of precursor and sequence ions was set to 10 ppm and 0.35 Da, respectively. Oxidation (M), propionamide (C), acetylation (protein N-terminus) and phosphorylation (STY) were used as variable modifications. A significance threshold of 0.05 was used based on decoy database searches and a peptide ion score cut-off of 20 was applied. In addition, tandem mass spectra of phosphopeptides were manually verified. Phosphorylation degrees for each residue were estimated by manually comparing relative MS peak intensities in the extracted ion chromatograms of the corresponding peptide–phosphopeptide pairs as described (Boehm et al., 2012; Seidler et al., 2009).

Protein masses were analyzed by matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF-MS) using an Ultraflex-II TOF/TOF instrument (Bruker Daltonics, Bremen, Germany) equipped with a 200 Hz solid-state Smart beam laser. The mass spectrometer was operated in the positive linear mode. MS spectra were acquired over an m/z range of 3000–20,000 and data were analyzed using FlexAnalysis 2.4. software provided with the instrument.

Sinapinic acid was used as the matrix (1:2 ratio of saturated solution in acetonitrile to 0.1% trifluoroacetic acid) and samples were spotted undiluted using the dried-droplet technique. Where necessary, samples were diluted in TA₃₃ (33% acetonitrile, 0.1% trifluoroacetic acid in water). External calibration was performed using the Bruker Protein Calibration Standard I (Bruker Daltonics, Bremen, Germany).

Membranes with the spotted peptides (Table S2) were washed once with 100% ethanol for 10 min and three times with PBS+1 mM DTT. The membranes were blocked with 5% BSA in PBS+1 mM DTT for 3 h. After three additional washing steps, the membranes were incubated overnight with GST-SRPK79D (20 µg/ml) and GST (6 µg/ml) in PBS+1 mM DTT plus 5% BSA. The membranes were then washed three times with PBST+1 mM DTT and subsequently incubated for 1 h with 1:1000 anti-GST-Z5 antibody (catalog no. sc-459, Santa Cruz Biotechnology) in PBS plus 5% BSA. After incubation, the membranes were washed again three times with PBS+1 mM DTT and were then incubated with 1:5000 horseradish peroxidase (HRP)-coupled anti-rabbit antibody (catalog no. sc-2004, Santa Cruz Biotechnology) for 1 h. Final detection was done by adding electrochemiluminescence solution (PJK GmbH) after three more washing steps with PBS+1 mM DTT. Luminescence was detected by sequential scanning using an Intas Advanced Fluorescence Imager.

Immunohistochemistry was performed according to our standard protocol (Andlauer et al., 2014). Conventional confocal and STED images were acquired with TCS SP8 and TCS SP8 gSTED 3× microscopes (Leica Microsystems, Wetzlar, Germany), respectively. Images of fixed samples were acquired at room temperature. NMJ z-stacks had a step size of 0.3 μm between single optical slices. Images were acquired from third instar larval axons. All images were acquired using the LCS AF software (Leica Microsystems). For previous descriptions see Fouquet et al. (2009).

Dissections were performed in hemolymph-like HL3 saline (HL3; Stewart et al. 1994) by opening the larvae or embryo dorsally along the midline and removing the innards to grant visual access to the larval CNS axons. Dissections were fixed with 4% paraformaldehyde in PBS (pH 7.2) for 10 min. After fixation, the filets were washed with PBS plus 0.05% Triton X-100 (PBT) and blocked for 60 min in 5% normal goat serum (NGS). For immunostainings, the larvae were incubated with primary antibodies (see below) at 4°C overnight and subsequently washed in a 0.05% PBT solution for 2 h at room temperature. Larvae were then incubated overnight with secondary antibodies (see below) at 4°C. Washing procedures were repeated. Immunocytochemistry was equal for both conventional confocal and STED microscopy. Larvae were finally mounted either in Vectashield (Vector Laboratories) or Mowiol (Sigma-Aldrich). Immunofluorescence images were recorded from axon bundles of motor and sensory neurons that emanate proximal to the center of the larval ventral nerve cord. Antibody dilutions were: 1:100–1:200 mouse anti-NC82 (catalog no. nc82, Developmental Studies Hybridoma Bank); 1:500 rabbit anti-RIM-BP (Liu et al., 2011); 1:500 guinea pig anti-Unc13A (Böhme et al., 2016) and 1:250 Alexa Fluor 647 AffiniPure goat anti-HRP (catalog no. 123-605-021, Jackson ImmunoResearch). Confocal secondary antibodies [goat anti-mouse Alexa 488 (catalog no. ab150117, Abcam); goat anti-rabbit Cy3 (catalog no. ab6939, Abcam)] were diluted 1:500. Secondary antibodies used for STED imaging [goat anti-M-STAR635p (catalog no. 2-0002-007-5, Abberior); goat anti-Rb-Atto594 (catalog no. A11037, Invitrogen) and goat anti-guinea pig Atto594 (catalog no. A11076, Invitrogen)] were diluted 1:200.

Briefly, the signal of an HRP-Cy5 antibody was used as template for a mask, restricting the quantified area to the shape of the axon or nerve bundle. The original confocal stacks were converted to maximal projections. After background subtraction, a mask of the axonal area was created by applying a threshold to remove spurious low-intensity pixels. The segmentation of single spots was done semi-automatically via the ‘Find Maxima’ routine in ImageJ and by hand with the pencil tool and a line thickness of 1 pixel. To remove high-frequency noise, a Gaussian blur filter (0.5-pixel Sigma radius) was applied. The processed picture was then transformed into a binary mask using the same lower threshold value as in the first step. This binary mask was then projected onto the original unmodified image using the ‘min’ operation from the ImageJ image calculator. The axonal spots of the resulting images were counted with the help of the ‘analyze particle’ function with a lower threshold set to 1. The spot density was obtained by normalizing the total number of analyzed particles to the axonal area measured via horseradish peroxidase (HRP).

Conventional embedding was performed as described previously (Fouquet et al., 2009).

Unless otherwise stated, data were analyzed with GraphPad Prism 5 software using a one-way ANOVA with Tukey's post hoc test.

We thank Christine Quentin for excellent technical assistance, Susanne Schoch-McGovern, Universität Bonn, for providing a mSRPK2 plasmid, and Gourisankar Ghosh, University of California, San Diego, for providing SRPK1 and SRPK1^CoreΔlinker (SRPK1ΔNS1) expression constructs. We acknowledge access to the Core Facility BioSupraMol supported by the Deutsche Forschungsgemeinschaft.