Structure of Slitrk2–PTPδ complex reveals mechanisms for splicing-dependent trans-synaptic adhesion

Selective binding between pre- and postsynaptic adhesion molecules can induce synaptic differentiation. Here we report the crystal structure of a synaptogenic trans-synaptic adhesion complex between Slit and Trk-like family member 2 (Slitrk2) and receptor protein tyrosine phosphatase (RPTP) δ. The structure and site-directed mutational analysis revealed the structural basis of splicing-dependent adhesion between Slitrks and type IIa RPTPs for inducing synaptic differentiation.

Selective binding between the type IIa RPTPs and their cognate postsynaptic partners is regulated in part by alternative splicing of the type IIa RPTPs 3,4,6 . PTPd expressed in the developing brain consists of three immunoglobulin-like (Ig) and four fibronectin type III (Fn) domains in the extracellular region, where two splice sites exist within Ig2 and between Ig2 and Ig3 (Fig. 1a). Peptide insertions (termed mini-exon peptides) at these two splice sites (referred hereafter to as meA and meB) generate splice variants of PTPd. In this study, we determined the crystal structure between the first leucine-rich repeat (LRR1) of mouse Slitrk2 and the Ig1-Fn1 domains of PTPd containing both meA and meB. The structure showed that Slitrk2 LRR1 directly recognizes meB but not meA. Further structure-based mutational analyses using surface-plasmon resonance (SPR) spectroscopy and synaptogenic co-culture assay demonstrated that binding of Slitrk2 to PTPd depends on meB.

Results and Discussion
Overall structure. For structural studies, we initially examined the PTPd-binding region of Slitrk4 and found that mouse Slitrk4 LRR1 is sufficient for binding to the extracellular domain of PTPd (PTPd-ECD) (Supplementary Figure 1a). However, our co-crystallization trials using Slitrk4 LRR1 and the full-length or truncated PTPd-ECD failed. We then screened other Slitrk members in terms of their expression level in FreeStyle293F cells and their binding activities to PTPd-ECD and selected mouse Slitrk2 LRR1 as the next candidate for crystallization. After optimization of the length of PTPd-ECD (Supplementary Figure 1b), we finally determined the crystal structure of a complex between Slitrk2 LRR1 and PTPd Ig1-Fn1 at 3.35 Å resolution (Fig. 1a,b, Supplementary Figure 2a and Table 1). The asymmetric unit contains one complex, where Slitrk2 LRR1 binds to PTPd Ig1-Fn1 at a ratio of 151 (Fig. 1b, Supplementary Figure 2b). Slitrk2 LRR1 interacts with PTPd Ig2-3. The overall structure of PTPd Ig1-Fn1 exhibits an elongated shape. PTPd Ig1-2 forms a compact V-shaped unit, similarly to LAR or PTPs Ig1-2 15 . Ig3 is spatially separated from Ig1-2 by meB. The following Fn1 is linearly aligned to Ig3. Slitrk2 LRR1 comprises a central LRR with seven parallel b-strands flanked by N-and C-terminal caps, which are stabilized by disulfide bonds (Fig. 1c). The total nine b-strands in the N-terminal cap and LRR form a concave surface of Slitrk2. The interior of the convex side is rich in completely conserved phenylalanine residues, which form a ''Phe spine'' structure as observed in the Nogo receptor 16 (Fig. 1c and Supplementary Figure 3).
Binding interface. The concave surface of Slitrk2 LRR1 surrounds two strands in PTPd Ig2 and meB ( 234 ELRE 237 ) with a buried surface area of 644 Å 2 (Fig. 2a). Asp167 and Glu215 of Slitrk2 hydrogen bond with Arg236 of PTPd (the third residue of meB) (Fig. 2b). In addition, Asp142, Asn166 and Asp187 of Slitrk2 hydrogen bond with Arg233 of PTPd (one-residue upstream of meB). Surrounding these interactions, Arg114 and Arg189 of Slitrk2 hydrogen bond with Gln209 and Glu145 of PTPd, respectively, whereas Arg114 and His185 of Slitrk2 hydrogen bond with the main-chain O atoms of PTPd Val232 and Leu141, respectively. Furthermore, Tyr138 of Slitrk2 is stacked with Tyr231 of PTPd. These Slitrk2 residues involved in the interaction with PTPd Ig2-meB are mostly   conserved in all mouse Slitrk members (Supplementary Figure 3), suggesting that they can potentially recognize PTPd Ig2-meB in the same manner as Slitrk2. Slitrk2 LRR1 also interacts with PTPd Ig3 with a buried surface area of 450 Å 2 (Fig. 2c). Phe247 and His250 of Slitrk2 hydrophobically interact with Tyr273 of PTPd, which hydrogen bonds with the main-chain O atom of Slitrk2 Glu244 (Fig. 2d). Phe247 of Slitrk2 also hydrophobically interacts with Met289 of PTPd and appears to play a central role in the Slitrk2 LRR1-PTPd Ig3 interface. However, this phenylalanine residue is replaced by Thr, Ser or Pro in Slitrk1, 4 or 6, respectively (Supplementary Figure 3), which are inadequate to form hydrophobic interactions with Tyr273 and Met289 of PTPd. In addition, His250 of Slitrk2 is not conserved in Slitrk1, 4 or 6 (Supplementary Figure 3). Therefore, the observed Slitrk2-PTPd Ig3 interactions may be specific to Slitrk2, 3 and 5.
Splicing-dependent interactions and synaptogenic activity. In the crystal structure of the Slitrk2-PTPd complex, Slitrk2 LRR1 directly recognizes the meB insertion of PTPd, whereas the meA insertion is distant from Slitrk2 LRR1 (Fig. 2a). Therefore, the present structure clearly indicates that binding of PTPd to Slitrk2 depends on meB but not on meA. To ensure this finding, we examined the binding of Slitrk2 LRR1 to the PTPd variants containing either or both meA and meB (meA9B-, meA-B1 and meA9B1) by surface-plasmon resonance (SPR) spectroscopy (Fig. 3a). As expected, the meA9B1 and meA-B1 variants bound to Slitrk2 LRR1 with similar affinities, whereas the meA9B-variant hardly bound to Slitrk2 LRR1 (Fig. 3a, c and Supplementary Figure 4). We next assessed the observed Slitrk2-PTPd interactions by site-directed mutagenesis ( Fig. 3 and Supplementary Figure 4). The R236E mutation of PTPd and the D167A and E215A mutations of Slitrk2, which disrupt the meBspecific interactions, abolished the binding between Slitrk2 and PTPd. The D187A mutation of Slitrk2, which disrupts the hydrogen bond with Arg233 of PTPd (one-residue upstream of meB), also abolished the binding. Accordingly, the D167A, D187A or E215A mutant of Slitrk2 abolished or significantly reduced their synaptogenic activities in our co-culture assay (Fig. 4). On the other hand, the R114A mutation of Slitrk2 (PTPd Ig2-mediated interface), the Y273A mutation of PTPd or the F247A H250A double mutation of Slitrk2 (PTPd Ig3-mediated interface) hardly affected the binding. Therefore, the R114A or F247A H250A mutant of Slitrk2 exhibited the synaptogenic activity comparable to wild-type Slitrk2 (Fig. 4). These results perfectly support the finding that binding of PTPd to Slitrk2 depends on meB ( Fig. 3 and Supplementary Figure 4).

Structural comparison with the Slitrk1 LRR1-PTPd complex.
Recently, the crystal structure of the complex between Slitrk1 LRR1 and PTPd Ig1-3 has been reported 17 Figure 5d). Concomitantly, the PTPd Ig3mediated interfaces are different between the Slitrk1 and 2 complexes (Supplementary Figure 5e). Physiological significance of this difference remains obscure, because the PTPd Ig3-mediated interface appears unlikely to contribute to the affinity to Slitrk proteins as mentioned above.
Trans-synaptic clustering of the Slitrk1-PTPd complex has been proposed on the basis of the lateral interaction between neighboring complexes in the crystal 17 . This lateral interaction involves Arg72, Phe74 and Arg143 of Slitrk1 and Glu237 and Glu286 of PTPd. On the other hand, no higher-order clustering was observed in the crystal of the Slitrk2-PTPd complex (Supplementary Figure 2b). Arg143 of Slitrk1 corresponds to Ser147 of Slitrk2, which is exposed to the solvent (Supplementary Figure 5f). Arg72 and Phe74 of Slitrk1 correspond to Arg76 and Tyr78 of Slitrk2, respectively, which interact with Asp378 of the neighboring PTPd Fn1 but appear unable to form higher-order clustering (Supplementary Figure 5f).
In conclusion, we revealed that binding of Slitrk2 to PTPd depends on meB but not on meA, based on the crystal structure and mutational studies at the molecular and cellular levels. Our previous analysis of cDNA from the developing mouse brain showed that PTPs and LAR also have the splicing variants containing a four-residue insert as meB with the conserved third arginine residue 4 , suggesting that the meB-containing variants of all type IIa RPTPs can potentially bind to all Slitrk proteins. In fact, a recent study showed that the meB-containing variant of LAR can bind to Slitrk1 (ref. 17). Other example of splicing-dependent regulation of trans-synaptic adhesion through the type IIa RPTPs is the adhesion between PTPd and IL1RAPL1 or IL-1RAcP 3,4 . Further studies for these complexes, together with the present study, will lead to complete understanding of splicing-dependent regulation of trans-synaptic adhesions for synaptic differentiation.

Methods
Protein expression and purification. The gene encoding mouse PTPd Ig1-Fn1 (residues 28-418) was amplified from cDNA (accession No. NM_011211.3) by PCR and cloned into pEBMulti-Neo vector (Wako Pure Chemical Industries) with the Nterminal signal sequence derived from pHLsec vector 18 23 . Data collection and refinement statistics were summarized in Table 1. The buried surface area was calculated using the program PISA 24 . All structural figures are prepared using the program PyMol (Schrödinger, LLC). SPR analysis. SPR experiments were carried out by using Biacore T200 (GE healthcare) at 25uC in 10 mM HEPES-Na buffer (pH 7.9) containing 150 mM NaCl and 0.05% Tween-20. The wild-type or mutant Slitrk2 LRR1 (D167A, D187A, E215A, R114A or F247A H250A) was immobilized on a CM5 sensor tip by the aminecoupling method. The splicing variant (meA9B1, meA9B-or meA-B1) or mutant (R236E or Y273A) of PTPd Ig1-Fn1 was injected at concentrations ranging from 15.625 to 4,000 nM. 10 mM NaOH was used as a regeneration buffer.
Synaptogenic assay. Primary cortical cultures were prepared from mice at E18 essentially as described previously 4 . Expression vectors for the mutated forms of Slitrk2-LRR1-Fc were generated by PCR-based mutagenesis using pEB6-Slitrk2-LRR1-Fc as a template. Fc and the mutated forms of Slitrk2-LRR1-Fc in FreeStyle293F cell culture medium were bound to Protein A-conjugated magnetic particles (smooth surface, 4.0-4.5 mm diameter; Spherotech). Beads coupled with Fc or the Fc fusion proteins were added to cortical neurons at days in vitro 16. After 24 hours, cultures were fixed for immunostaining with mouse anti-Bassoon antibody (Stressgen, 15400) and rabbit anti-human Fc (Rockland, 151000).
Image acquisition and quantification. Images of bead-neuron co-cultures were collected from at least two separate experiments. Image acquisition and quantification for the co-culture assays were performed essentially as previously described 4 . Briefly, the intensities of immunostaining signals for Bassoon were measured as the optical mean density within a circle of 7-mm diameter enclosing coated-bead. Statistical significance was evaluated by one-way ANOVA followed by post hoc Tukey's test.