Structural basis for extracellular cis and trans RPTP[sigma] signal competition in synaptogenesis

Abstract

Receptor protein tyrosine phosphatase sigma (RPTPσ) regulates neuronal extension and acts as a presynaptic nexus for multiple protein and proteoglycan interactions during synaptogenesis. Unknown mechanisms govern the shift in RPTPσ function, from outgrowth promotion to synaptic organization. Here, we report crystallographic, electron microscopic and small-angle X-ray scattering analyses, which reveal sufficient inter-domain flexibility in the RPTPσ extracellular region for interaction with both cis (same cell) and trans (opposite cell) ligands. Crystal structures of RPTPσ bound to its postsynaptic ligand TrkC detail an interaction surface partially overlapping the glycosaminoglycan-binding site. Accordingly, heparan sulphate and heparin oligomers compete with TrkC for RPTPσ binding in vitro and disrupt TrkC-dependent synaptic differentiation in neuronal co-culture assays. We propose that transient RPTPσ ectodomain emergence from the presynaptic proteoglycan layer allows capture by TrkC to form a trans-synaptic complex, the consequent reduction in RPTPσ flexibility potentiating interactions with additional ligands to orchestrate excitatory synapse formation.

ARTICLE

Received 30 Apr 2014 | Accepted 9 Sep 2014 | Published 11 Nov 2014

DOI: 10.1038/ncomms6209 OPEN

Structural basis for extracellular cis and trans RPTPs signal competition in synaptogenesis

Charlotte H. Coles1,w, Nikolaos Mitakidis1, Peng Zhang2, Jonathan Elegheert1, Weixian Lu1,Andrew W. Stoker3, Terunaga Nakagawa4, Ann Marie Craig2, E. Yvonne Jones1 & A. Radu Aricescu1

Receptor protein tyrosine phosphatase sigma (RPTPs) regulates neuronal extension and acts as a presynaptic nexus for multiple protein and proteoglycan interactions during synaptogenesis. Unknown mechanisms govern the shift in RPTPs function, from outgrowth promotion to synaptic organization. Here, we report crystallographic, electron microscopic and small-angle X-ray scattering analyses, which reveal sufcient inter-domain exibility in the RPTPs extracellular region for interaction with both cis (same cell) and trans (opposite cell) ligands. Crystal structures of RPTPs bound to its postsynaptic ligand TrkC detail an interaction surface partially overlapping the glycosaminoglycan-binding site. Accordingly, heparan sulphate and heparin oligomers compete with TrkC for RPTPs binding in vitro and disrupt TrkC-dependent synaptic differentiation in neuronal co-culture assays. We propose that transient RPTPs ectodomain emergence from the presynaptic proteoglycan layer allows capture by TrkC to form a trans-synaptic complex, the consequent reduction in RPTPs exibility potentiating interactions with additional ligands to orchestrate excitatory synapse formation.

1 Division of Structural Biology, Wellcome Trust Centre for Human Genetics, University of Oxford, Roosevelt Drive, Oxford OX3 7BN, UK. 2 Brain Research Centre and Department of Psychiatry, University of British Columbia, Vancouver, British Columbia, Canada V6T 2B5. 3 Cancer Section, Institute of Child Health, University College London, 30 Guilford Street, London WC1N 1EH, UK. 4 Department of Molecular Physiology and Biophysics, Vanderbilt University, School of Medicine, 702 Light Hall (0615), Nashville, Tennessee 37232-0615, USA. w Present address: Laboratory for Axon Growth and Regeneration, German

Center for Neurodegenerative Diseases (DZNE), Ludwig-Erhard-Allee 2, 53175 Bonn, Germany. Correspondence and requests for materials should be addressed to A.R.A. (email: mailto:[email protected]

Web End [email protected] ) or to E.Y.J. (email: mailto:[email protected]

Web End [email protected] )

NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209

Neuronal synaptogenesis is orchestrated by cell surface receptors, which dene the formation and functionality of distinct synapse classes1,2. These organizer molecules

can act as scaffolds or hubs, to integrate multiple inputs into a unied cellular response. This concept is well established for intracellular systems, but only now emerging for cell surface molecules such as neurexin-neuroligin and repulsive guidance molecule (RGM)-neogenin protein complexes35. At the axonal surface, type IIa receptor protein tyrosine phosphatases (RPTPs) (type IIa RPTPs, for example, RPTPs, RPTPd and leukocyte common antigen-related (LAR) in vertebrates, dLAR in Drosophila), are a recently identied nexus for extracellular interactions, receiving signals and transmitting them intracellularly to the cytoskeleton to regulate neuronal extension and guidance, as well as synaptic organization69.

Heparan sulphate proteoglycans (HSPGs) and chondroitin sulphate proteoglycans (CSPGs) play important roles in the modulation of RPTP signalling in the nervous system1015. At the Drosophila neuromuscular junction, the HSPGs dSyndecan and dDallylike opposingly regulate dLAR-mediated synaptic morpho-genesis and active zone function12. While the interaction of dLAR with presynaptic dSyndecan promotes bouton growth, postsynaptic dDallylike competes with dSyndecan for dLAR binding, leading to an inhibition of growth and active zone stabilization12. Interactions of the RPTPs ectodomain with

HSPGs and CSPGs modulate axonal growth both during development and post injury10,13,14,16. HSPGs cluster RPTPs, a characteristic proposed to drive a localized imbalance of protein tyrosine phosphorylation and hence promote growth14. Once the axon has reached its nal target, RPTPs can establish direct interactions with multiple postsynaptic proteins. In vertebrates, these include the TrkC receptor protein tyrosine kinase, Netrin-G

ligand-3 (NGL-3), interleukin-1 receptor accessory protein and Slit- and Trk-like receptors 1 and 2 (Slitrk1 and Slitrk2)1720. These trans-synaptic complexes mediate bi-directional excitatory synapse formation, simultaneously triggering presynaptic differentiation and an accumulation of synaptic vesicles, and clustering of the postsynaptic density7,9.

There are two major neuronal RPTPs isoforms, sharing a common intracellular catalytic region and an extracellular region predicted to contain three immunoglobulin (Ig)-like domains followed by either ve or nine bronectin (FN) type III domains, in the central and peripheral nervous systems, respectively21. Further isoforms include a combination of four mini-exons (meA-meD) that may modulate interactions with protein partners (Fig. 1a)8,21. Previous mutagenesis and structural studies have demonstrated that the proteoglycan-binding site lies on Ig1 of RPTPs, and comprises an extended positively charged surface of basic residues10,13,14. Binding of postsynaptic TrkC, is reported to require the N-terminal three Ig domains of RPTPs17; the NGL-3-binding site has been mapped to the FN1-2 domains18.

The properties that t the RPTPs ectodomain for its function as an integrative hub for signalling in synaptogenesis are unknown. Here we report a molecular level analysis of the RPTPs ectodomain and of its direct interactions with the postsynaptic binding partner TrkC. We reveal that the multi-domain extracellular region of RPTPs is unexpectedly exible.

This characteristic confers sufcient conformational freedom to allow its binding to both pre- or postsynaptic ligands. RPTPs:TrkC crystal structures provide an explanation for the specicity of this interaction and also highlight an overlap of TrkC and proteoglycan-binding sites on RPTPs. This observation suggests that there is competition between TrkC and heparan

meC

604

1,012

meA

meB

4 5 6 7

8 9 D1 D2

meD

1,882 C

1

GAG

N 1

30 321

602

2

3 1 2 3

SP

Ig domains FN3 domains Phosphatase domains

TM

Ig13 Ig1-FN3 sEcto

30 30

1,249

Ig1

Ig1

1 2 3 1 2 3

8 9

Ig2

Ig2

N

Ig1-3

Ig1-3

Ig3

Ig3

FN1

FN1

90

67

170

52

C

FN2

FN2

FN3

FN3

1 2 3 1 2 3

Figure 1 | RPTPr ectodomain exibility. (a) RPTPs domain organization. N, amino-terminus (extracellular); SP, secretion signal peptide;

TM, transmembrane; C, C terminus (intracellular); Ig, immunoglobulin-like; FN, bronectin type-III; GAG, glycosaminoglycan-binding site (lled arrowhead). Alternative splicing inserts: FN domains 47 and mini-exons AD (open arrowhead). (b) Ribbon and surface representations of the human RPTPs Ig1-FN3 crystal structure. N-linked glycans in atom representation. (c) Ig3 movement in Ig1-FN3 relative to Ig13 (grey, PDB ID: 2YD9) structure.

Representative RPTPs Ig1-FN3 (d) and RPTPs sEcto (e) negative-stain electron microscopy class averages. Scale bar, 10 nm. Full sets of RPTPs Ig1-FN3 and sEcto class averages are provided in Supplementary Figs 2 and 3.

2 NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209 ARTICLE

sulphate for RPTPs binding, and we provide support for this notion in biophysical and cellular assays. Overall, our study provides novel insights into the mechanisms determining the hierarchy and functional consequences of RPTPsligand interactions.

ResultsThe RPTPr ectodomain exhibits extensive exibility. We rst investigated the molecular characteristics of RPTPs that allow presentation of the N-terminal Ig domains for both pre- and postsynaptic ligand binding. We determined the 3.15 crystal structure of a six N-terminal domain human RPTPs construct (termed Ig1-FN3, Fig. 1a), which contains all the binding sites for synaptic ligands identied to date. This construct maintains the V-shaped Ig12 arrangement previously reported14,22, followed by an extended conformation of the sequential domains Ig3, FN1 and FN2 (Fig. 1b; Table 1). Superposition of Ig1-FN3 with the crystal structure of Ig13 (ref. 14) revealed a hinge point between domains Ig2 and Ig3 (Fig. 1c). The four amino-acid meB exon would extend this apparently exible linker further, by four residues (Supplementary Fig. 1). While FN1 and FN2 domains align approximately with the long axis of the molecule, the C-terminal FN3 domain folds back, suggesting the FN2-3 linker may also be a exion point. Since no substantial Ig3-FN1, FN1FN2 or FN2FN3 inter-domain interfaces are apparent, we hypothesized that when released from crystal packing constraints each of the Ig2-FN3 region linkers may provide substantial exibility. To test

this, single-particle negative-stain electron microscopy (EM) class averages of human RPTPs Ig1-FN3 were calculated. These showed a broad range of conformations (Fig. 1d; Supplementary Fig. 2). This exibility, reminiscent of hinge points in aneurexin23, is in marked contrast to the rigidity of homophilic cell adhesion molecules of similar size and domain organization, such as cadherins, RPTPm and SYG-1/SYG-2 (refs 2426).

We extended our EM analysis to the full ectodomain of the eight-domain isoform of human RPTPs (sEcto; Fig. 1a,e;

Supplementary Fig. 3). The 150 class averages generated reinforce our conclusions from the six-domain structural analyses. The RPTPs ectodomain exhibits a surprisingly large exibility, with observed conformations ranging from almost fully extended, to essentially bent double (Fig. 1e; Supplementary Fig. 3). To control for the potential risk of artefacts associated with negative staining, we also performed small-angle X-ray scattering (SAXS) measurements, at a physiological pH (7.4), for both human RPTPs Ig1-

FN3 and sEcto. This analysis further supports the observation that both proteins are exible and are likely to adopt multiple conformations in solution (Supplementary Fig. 4). Taken together, crystallographic, EM and SAXS analyses demonstrate that the ectodomain of RPTPs is able to explore a large conformational space.

Structural analysis of the RPTPr:TrkC trans-synaptic complex. How do these conformational properties contribute to the interaction of RPTPs with ligands? We sought to compare and

Table 1 | Data collection and renement statistics.

hRPTPr Ig1-FN3 cRPTPr Ig1-2 cTrkC LRRIg1

cryst cRPTPr Ig1-3 cTrkC LRRIg12Q

Data collectionSpace group P6122 P2 P1Cell dimensionsa, b, c () 198.8, 198.8, 132.4 68.3, 122.2, 98.6 84.4, 93.1, 99.4 a, b, g () 90.0, 90.0, 120.0 90.0, 109.8, 90.0 73.4, 89.5, 74.2

Resolution () 99.403.15 (3.233.15)* 63.962.50 (2.562.50) 81.023.05 (3.133.05) Rmerge 7.8 (99.1) 19.2 (142.4) 6.9 (34.7)

Rpimw 3.2 (41.0) 5.7 (57.2) 5.6 (60.3) CC1/2z 99.8 (63.9) 99.5 (58.4) 99.8 (67.2)

I/sI 17.8 (2.1) 10.7 (1.6) 8.8 (1.5) Completeness (%) 95.6 (96.1) 99.8 (99.3) 96.3 (96.6)

Redundancy 7.6 (7.5) 11.6 (7.1) 1.8 (1.8)

RenementResolution () 99.403.15 (3.233.15) 63.962.50 (2.562.50) 81.023.05 (3.133.05) No. reections 25,619 (1,858) 52,652 (3,833) 51,063 (3,823) Rwork/Rfree 23.4 (37.1)/26.5 (37.7) 20.9 (32.2)/24.7 (36.2) 22.6 (37.2)/24.0 (38.6)

No. of atomsProtein 4,380 8,536 10,630 NAGs 2 5 1 SO42 ions 2

Water 118

B-factorsProtein 122.2 57.2 116.4 NAGs 153.0 83.0 176.3 SO42 ions 90.9

Water 45.0

R.m.s.d.Bond lengths () 0.005 0.008 0.006 Bond angles () 1.008 1.243 1.045

RPTP, Receptor protein tyrosine phosphatase; r.m.s.d., root mean squared deviation.*Values in parentheses are for highest-resolution shell. Each structure is based on a single crystal.

wR (the precision-indicating merging R-value) 1/(N 1) R , where N is the redundancy. zCC is the mean intensity correlation coefcient of half-data sets53.

NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209

contrast complex formation between the RPTPs N-terminal region and axonal HSPGs14 with the trans-synaptic interaction between RPTPs and TrkC. The TrkC ectodomain comprises an

N-terminal leucine-rich repeat (LRR) domain, followed by two Ig domains (Fig. 2a). Equilibrium surface plasmon resonance (SPR) assays conrmed previously reported data17 that the minimal units required for full afnity binding are RPTPs Ig13 and TrkC

LRRIg1 (Supplementary Fig. 5a,b; Supplementary Table 1). RPTPs Ig12 does retain TrkC binding, although with an approximately vefold-reduced afnity compared with RPTPs Ig13 (Kd 2.4 mM versus 551 nM, Supplementary Fig. 5a). TrkC

LRRIg1 had also previously been shown to be the minimal unit required for the synaptogenic activity of TrkC17. To facilitate crystallization, a chicken TrkC construct (TrkC LRRIg1cryst)

was generated, which removed putative sites of N-linked glycosylation and residues 6377, a predicted disordered loop (Supplementary Fig. 6). A 2.5 crystal structure of a chicken RPTPs Ig12:TrkC LRRIg1cryst complex was determined, revealing a 1:1 stoichiometry (Fig. 2b; Table 1; Supplementary Fig. 5c), in agreement with results from multi-angle light

scattering (MALS) analysis (Supplementary Fig. 5d). The structure is consistent with a trans RPTPs:TrkC complex spanning the synaptic cleft (Fig. 2b). The V-shaped RPTPs Ig12 module contacts an extended TrkC surface consisting of the LRR convex face and Ig1 domain, with a buried surface area of 1,093 2 per molecule (Fig. 2b).

Three major contact sites constitute the proteinprotein interface in the complex crystal structure: site 1, RPTPs Ig1:TrkC

Ig1 (Fig. 2c; Supplementary Fig. 7); site 2, RPTPs Ig1:TrkC LRRIg1 inter-domain region (Fig. 2d; Supplementary Fig. 7) and site 3, RPTPs Ig2:TrkC LRR (Fig. 2e; Supplementary Fig. 7).

Electrostatic interactions involving RPTPs residues R96 and R99 and TrkC residues D240 and D242 dominate interactions at site 1 (Fig. 2c). Intriguingly, R96 and R99 form part of the extended positively charged surface on RPTPs Ig1 (ref. 14) and are absolutely required for RPTPs interactions with HSPGs10,14, suggesting that TrkC and proteoglycans may compete for binding to RPTPs. At site 2, Q75 in RPTPs, interacts with

TrkC residues E287 and Q148, while the side chains of RPTPs E78 and TrkC R121 form a salt bridge (Fig. 2d). Site 3 centres on

63 77

TrkC TM 32 463

GSSNGNTSINITDIS

1 612

LRRIg1cryst 32 302

1 2 3 C 1

N N C

2

SP

TM

LRR domain

Ig domains

LRRIg12N 32 302

T97

T97

LRRIg2 32 398

H254

H254 V241

V241

D242

R99 D242

R99

R96

R96

Lys

Lys loop

loop

D240

D240

C

Ig1

Ig1

S74

S74

Ig1

Ig1

E287

E287

N

LRR

LRR

Ig2

Ig2

Q75

Q75

R121

R121 Q148

Q148

C N

E78

E78

55 20

C

C

K203

K203

E100

E100

N221

N221

V97

V97

Y223

Y223

N

* TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrkkkkkkkkkkk

kC

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

N

Figure 2 | Trans-synaptic RPTPr:TrkC complex crystal structure. (a) TrkCTK- (non-catalytic isoform) domain organization. LRR, leucine-rich repeat region (N, N-terminal cysteine-rich region; 13, leucine-rich repeats; C, C-terminal cysteine-rich region). Putative N-linked glycosylation sites, lollipops;

lled lollipops remain in LRRIg1cryst construct. (b) Space-lled and tube representations of chicken RPTPs Ig12:TrkC LRRIg1cryst crystal structure. N-linked glycans in atom representation. Disordered RPTPs Lys-loop, blue dotted line; TrkC LRRIg1cryst amino-acid residue 6278 junction, asterisk.

(ce) Detailed view of bonding interactions at RPTPs:TrkC interface for binding sites 13. Corresponding electron density is illustrated in Supplementary Fig. 7. Potential electrostatic and hydrogen bonds, black dashed lines; oxygen atoms, red; nitrogen atoms, bluewhite.

4 NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209 ARTICLE

a K203E100 salt bridge between RPTPs and TrkC residues, respectively (Fig. 2e).

These three interaction sites are consistent with RPTPs binding TrkC, but not TrkA or TrkB17; of the six predominantly charged (D240, D242, E287, Q148, R121 and E100) TrkC residues observed to have side-chain-mediated RPTPs interactions, only E287 is conserved across the other

Trk family members (Supplementary Fig. 6). Although the key RPTPs residues discussed above are conserved in human RPTPd and LAR (Supplementary Fig. 1), the specicity of TrkC binding for RPTPs can be rationalized through closer inspection of type

IIa RPTP sequence alignments and the chicken RPTPs Ig1 2:TrkC LRRIg1cryst crystal structure (Fig. 3a). At site 1, we hypothesized that substitution of P98 (in RPTPs) to H98 in RPTP LAR, would result in a movement of the carbonyl group of T97 to ease the cis-conformation of residue 98 to trans (see PDB ID: 2YD5), disrupting interaction with the TrkC H254 carbonyl group. At site 2, we anticipated that substitution of S74 (in RPTPs) to N74 in RPTPd would result in the loss of this residues interaction with TrkC D240. To validate these predictions, RPTPs Ig13 P97V T98H (LAR-like Ig13) and

N73S S74N (RPTPd-like Ig13) proteins were generated, which

indeed displayed LAR- and RPTPd-like binding to TrkC in SPR analyses (Fig. 3b,c; Supplementary Fig. 8a,b; Supplementary Table 1).

Validation of the RPTPr:TrkC binding mode. To conrm the contribution of the RPTPs:TrkC interaction sites, we introduced point mutations into the interfaces on either protein (Fig. 4a) and measured the resulting dissociation constants (Kd) using SPR. As predicted, mutations in either TrkC (D240A D242A) or

RPTPs (R96A R99A, affecting the glycosaminoglycan (GAG)-

binding arginine residues), completely abolished both human and

chicken RPTPs binding to mouse and chicken TrkC, respectively

(Fig. 4b,c; Supplementary Fig. 8cf; Supplementary Table 1). The RPTPs Y223S mutation was designed to disrupt binding at site 3 by introducing an N-linked glycosylation site at N221, and RPTPs:TrkC binding was indeed ablated (Fig. 4c; Supplementary

Fig. 8e,f; Supplementary Table 1). In agreement with the structural and biophysical data, TrkC transmembrane (TrkC TM) D240A D242A expressing COS-7 cells, unlike wild-type TrkC

TM expressing positive controls, were unable to induce pre-synaptic differentiation in co-cultured rat hippocampal neurons despite comparable levels of cell surface expression (Fig. 4d; Supplementary Fig. 9).

We had noted that the interaction afnity between engineered proteins used for RPTPs:TrkC complex crystallization (chicken

RPTPs Ig13 and TrkC LRRIg1cryst) was some 20-fold lower

following the TrkC 6377 loop deletion (Kd 4.8 mM versus 216 nM; Fig. 5a; Supplementary Fig. 8g; Supplementary Table 1). We therefore explored the possible contribution of the 6377 loop to the RPTPs:TrkC interaction by engineering a construct termed TrkC LRRIg12Q, comprising the full sequence but still removing two predicted N-linked glycosylation sites by introducing N68Q and N72Q mutations (Supplementary Fig. 6). This construct did indeed bind RPTPs with enhanced afnity (Fig. 5a;

Supplementary Fig. 8g,h; Supplementary Table 1), while TrkC TM2Q induced comparable levels of presynaptic differentiation in co-cultured rat hippocampal neurons to wild-type TrkC TM (Fig. 4d). We determined the 3.05 crystal structure of this chicken RPTPs Ig13:TrkC LRRIg12Q complex in an attempt to visualize this additional, fourth, binding site (Table 1). Crystals grew in a new space group, P1, with three RPTPs:TrkC complexes in the asymmetric unit (a.s.u.). These align very closely with the two complexes/a.s.u. observed in the previous P2 structure (Fig. 5b; root mean squared deviation between 446 equivalent Ca residues of the P2 complex1 relative to P2 complex2 and P1 complexes1-3 was calculated to be 0.82 ,0.83 , 0.80 and 0.70 , respectively), providing additional

97

98

74

200

Response (RU) Response (RU)

100

LAR Ecto

LAR-like Ig13

0

5 10 15 20

Arg loop

Lys loop

73

Conc (M)

RPTP Ecto

RPTP Ecto

RPTP-like Ig13

RPTP v LAR

RPTP v RPTP

RPTP Ig13

800

H254

H254

E287

E287

600

Q148

Q148

R99

R99

R96

R96

400

D240

D240

N73

N73

Q75

Q75 S74

S74

200

T97

T97

0

10 20 30 40

P98

P98

Conc (M)

Figure 3 | TrkC binding preferences for type IIa RPTP family members. (a) Type IIa RPTP sequence alignments and detailed views of the RPTPs:TrkC crystal structure at binding site 1 (left) and binding site 2 (right). Blue boxes indicate RPTPs residues required for proteoglycan binding, and redboxes highlight key sequence differences conferring TrkC-binding specicity. Colour scheme as in Fig. 2: dark blue, RPTPs Ig1; pink, TrkC LRR; magenta,

TrkC Ig1. (b) SPR analysis of human type IIa RPTP ectodomains binding to immobilized mouse TrkC LRRIg1. Measured binding values: RPTPs Ecto,Kd 516 nM and Bmax 217 RU; RPTPd Ecto, Kd422 mM and Bmax4433 RU; RPTP LAR, Kd and Bmax not determined. (c) SPR analysis of chicken TrkC LRRIg1

binding to immobilized chicken RPTPs Ig13, RPTPs N73S S74N (RPTPd-like) Ig13 and RPTPs P97V T98H (LAR-like) Ig13. Measured binding

values: RPTPs Ig13, Kd 3.5 mM and Bmax 837 RU; RPTPd-like Ig13, Kd421 mM and Bmax4674 RU; RPTP LAR-like Ig13, Kd and Bmax not determined. For

sensograms see Supplementary Fig. 8a,b.

NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209

Ig1

Ig1

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

RP

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

PTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTP

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P[afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846]

600

Ig1

Ig1

N R

C

Ig2

Ig2

Response (RU)

400

LRR

LRR

R96A+R99A

Y223S R227A+R228A

200

N

C

C

180 TrkC only

TrkC LRRIg1

RPTP Ig13 concentration (M)

TrkC LRRIg1 D240A+D242A

D240A+D242A

0

1

2

3

5 10 15

1

2

800

3

4

600

Response (RU)

400

R99

R96

D240D242

R121

*

E100

* E287

Ig13 WT

Ig13 Y223S

Ig13 R227A+R228A

E78

200

Ig13 R96A+R99A

K203

0

20 40

TrkC LRRIg1 concentration (M)

TrkC TM TrkC TM D240A+D242A TrkC TM2Q

100

TrkC TM

TrkC TM 2Q

TrkC TM D240A+D242A

Synapsin MAP2 mVenus Synapsin

Induced synapsin (au)

50

**

0

Figure 4 | Validation of RPTPr:TrkC binding interfaces. (a) Opened-view surface representation of the chicken RPTPs Ig12:TrkC LRRIg1cryst crystal

structure. RPTPs and TrkC interface residues are coloured grey and interface mutants used in biophysical and cellular assays are highlighted in black (middle panel). Binding sites 14 are labelled. RPTPs and TrkC are coloured by electrostatic potential (bottom panel) from red ( 8 kT/e) to blue

( 8 kT/e), illustrating complementary charged patches at binding sites 13 (note that the basic RPTPs Lys-loop residues 6871 are absent in the

RPTPs:TrkC complex crystallographic model). (b) SPR analysis of human RPTPs Ig13 binding to immobilized mouse TrkC LRRIg1 and TrkC LRRIg1 D240A D242A. Measured binding values: TrkC LRRIg1, Kd 258 nM and Bmax 540 RU; TrkC LRRIg1 D240A D242A, Kd and Bmax not determined.

(c) Mouse TrkC LRRIg1 binding to immobilized human RPTPs Ig13 WT, R96A R99A, Y223S and R227A R228A. Measured binding values: RPTPs Ig1

3 WT, Kd 1.8 mM and Bmax 802 RU; Ig13 R227A R228A, Kd 7 mM and Bmax 806 RU; Ig13 R96A R99A and Y223S, Kd and Bmax not determined.

For sensograms see Supplementary Fig. 8cf. (d) Induced synapsin clustering in rat hippocampal neurons by TrkC TM (WT), TrkC TM_D240A

D242A and TrkC TM2Q expressing COS-7 cells. Analysis of variance, Po0.0001; **Po0.001 compared with TrkC TM by post hoc Bonferronis multiple comparison test, n 26 cells from two experiments. Scale bar, 10 mM. Relative cell surface expression levels are shown in Supplementary Fig. 9.

support for the relevance of the observed RPTPs:TrkC interaction mode. The site 4 interface is not well resolved in this 3.05- resolution structure, although we do observe an additional TrkC helix (formed by residues 6980), which interacts with RPTPs

Ig2 predominantly via potential packing of the I73 side chain against a hydrophobic region consisting of RPTPs residues V144,

Y223 and TrkC L56, T74 and L101 (Fig. 5c,d). The RPTPs Ig2Ig3 linker (V226-A230) and the Ig3 domain also lack well-ordered electron density in this structure. However, a R227A

R228A double mutation reduced RPTPs:TrkC binding (Fig. 4c;

Supplementary Fig. 8e,f; Supplementary Table 1), as did insertion

of the meB mini-exon between R225 and V226 in SPR-binding assays, supporting the involvement of the RPTPs Ig2Ig3 linker in the auxilary binding site 4 (Supplementary Fig. 8i; Supplementary Table 1).

HSPGs compete with TrkC for RPTPr binding. The overlap of the TrkC- and GAG-binding sites on RPTPs, prompted us to investigate the notion that proteoglycan competition with TrkC has the potential to modulate RPTPs function in synaptogenesis.

Initially, we tested whether soluble HS or the HS-mimetic

6 NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209 ARTICLE

TrkC LRRIg12Q

TrkC LRRIg12N

TrkC LRRIg1cryst

600

Y223

Y223

Response (RU)

L101

L101

400

V144

V144

T74

T74

200

L56

L56

0

1 2 3 4 5

I73

I73

RPTP Ig13 concentration (M)

C

Ig1

N

N

Ig1

TrkC

T74

T74

S70

S70

I71

I71

D75

D75

Ig2

I73

I73

LRR

RPTP[afii9846]

P2 complex1

P2 complex1 P2 complex2

P2 complex2

T69

T69

I76

I76

P1 complex1

P1 complex1 P1 complex2

P1 complex2 P1 complex3

P1 complex3

2530

Q72

Q72

Figure 5 | Characterization of the potential accessory RPTPr:TrkC-binding site 4. (a) SPR analysis of chicken RPTPs Ig13 binding to immobilized chicken TrkC LRRIg1cryst, LRRIg12N and LRRIg12Q. Measured binding values: TrkC LRRIg1cryst, Kd 4.8 mM and Bmax 761 RU; LRRIg12N, Kd 216 nM and

Bmax 416 RU; LRRIg12Q, Kd 38 nM and Bmax 590 RU. For sensograms see Supplementary Fig. 8gh. (b) Alignment of the two RPTPs:TrkC complexes

observed in the chicken RPTPs Ig12:TrkC LRRIg1cryst (P2 space group) and three in the chicken RPTPs Ig13:TrkC LRRIg12Q (P1 space group) crystal

structures. The orientation of the structures is identical to Fig. 2b (lower panel). (c) Additional features observed in complex 1 from the RPTPs Ig13:TrkC LRRIg12Q crystal structure. Residues within the 6377 loop that were not present in the P2 crystal structure are coloured blue and the remaining missing residues are indicated by dotted lines. View rotated relative to b as indicated. TrkC LRR, magenta; RPTPs, cyan. (d) SigmaA weighted 2Fo Fc

electron density map (grey) contoured at 1s and carved at 2.2 around loop residues 6976.

heparin-dp10 could inhibit the interactions between TrkC and either wild-type RPTPs or a quadruple K67A K68A

K70A K71A mutant (RPTPs Ig13 DK), which precludes GAG

binding14 while retaining wild-type TrkC binding (Fig. 6; Supplementary Fig. 10a,b; Supplementary Table 1). Increasing concentrations of HS or heparin-dp10 were able to inhibit the binding of RPTPs Ig13 WT, but not RPTPs Ig13 DK to immobilized TrkC LRRIg1 (Fig. 6a; Supplementary Fig. 10a,b; Supplementary Table 1). To investigate the TrkC versus GAG competition in a cellular setting, we added heparin-dp10 to co-cultures of TrkC TM expressing COS-7 cells and rat hippocampal neurons. Induced presynaptic differentiation in the neurons decreased by twofold compared with mock-treated co-cultures upon heparin-dp10 addition (Fig. 6b,c). Furthermore, treatment of co-cultures with a mixture of heparinases I, II and III, to digest heparan sulphate GAGs, signicantly enhanced presynaptic induction by TrkC TM (Fig. 6b,c), suggesting that native HSPGs may limit synapse development through RPTPs:TrkC, by direct competition for binding (Fig. 6d). In contrast to TrkC, the interaction of RPTPs with another trans-synaptic protein ligand, NGL-3, reported to bind at the FN1-2 domains, appears insensitive to proteoglycans. Neither heparin-dp10 nor heparinase treatment affected NGL-3-induced synaptogenesis in the co-culture system (Fig. 6b,c), and RPTPs:NGL-3

binding in SPR assays showed no major reduction upon HS addition (Supplementary Fig. 10c,d; Supplementary Table 1).

DiscussionWhile all three vertebrate type IIa RPTP family members bind NGL-3 (refs 18,27) and interleukin-1 receptor accessory protein19, and both RPTPs and RPTPd bind to Slitrk1 and

Slitrk2 (refs 20,28), the IL-1 receptor accessory protein-like 1 (IL1RAPL1) interacts predominantly with RPTPd29,30 and the receptor protein tyrosine kinase TrkC interacts with RPTPs17. We therefore used this latter proteinprotein interaction as our exemplar for trans-synaptic RPTPs action. RPTPs and TrkC exhibit broad and overlapping expression patterns in the adult nervous system18,21,3133. Multiple ligand interactions and signalling pathways are disrupted in RPTPs- and TrkC-decient mice though, making assessment of any overlap in phenotypes difcult3437. A direct comparison of the effect of either TrkC knockdown17 or RPTPs knockout38,39 upon synapse structure and number in vivo, is similarly complicated by the parallel role of RPTPs in regulating axon sprouting. The detailed explanation of RPTPs:TrkC specicity that we offer in this study provides the information to enable the design of new experiments to dissect the precise contribution of this interaction to

NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209

**

100

150

0

RPTPbound (%)

Induced synapsin (a.u.)

100

50

*

50

cIg13, dp10, IC50 = 48 M

cIg13 K, dp10

hIg13, HS, IC50 = 55 M

hIg13 K, HS

0

Mock

dp10

Heps

Mock

dp10

Heps

106 105 104 10

Concentration GAG (M)

TrkC TM NGL-3

Mock dp10 Heparinases

SynapsinMAP2

Synapsin Synapsin MAP2 NGL-3-CFP

Synapsin

TrkC Tm mVenus

Trrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk

kC

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

Lys

Lys loop

loop

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

R

RP

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

PTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTP

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P

P[afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846][afii9846]

Ig1

Ig1

SOS

SOS Ig1

Ig1

LRR

LRR

Ig2

Ig2

Figure 6 | GAG-mediated inhibition of synaptic RPTPr:TrkC interaction and function. (a) SPR analysis of human RPTPs Ig13 and RPTPs Ig13 DK binding to immobilized mouse TrkC LRRIg1 in the presence of increasing concentrations of HS or heparin-dp10. For sensograms see Supplementary

Fig. 10a,b. (b,c) Induced synapsin clustering in rat hippocampal neurons by COS-7 cells expressing TrkC TM or NGL-3 in the presence of heparin-dp10, heparinases (Heps) or mock control. Analysis of variance Po0.0001, *Po0.01 and **Po0.001 compared with TrkC TM mock, whereas NGL-3 heparin-dp10 or heparainase groups were not signicantly different from NGL-3 mock by post hoc Bonferronis multiple comparison test, n 1626 cells

from two experiments. Scale bar, 10 mM. (d) Illustration of the partial overlap between GAG- and TrkC-binding sites on RPTPs. Top panel: the RPTPs:TrkC complex, rotated 180 around the y axis relative to Fig. 2b. Lower panel: sucrose octasulphate (SOS, grey/red) is modelled in the RPTPs GAG-binding site, an equivalent location to that observed in the LAR:SOS co-crystal structure (which is homologous with RPTPs, PDB ID: 2YD8). SOS (or GAG)

binding can out-compete the TrkC interaction with RPTPs.

synaptogenesis in vivo. The locations of RPTPs and TrkC are primarily reported as pre- and postsynaptic, respectively8,9, and we depict them as such in our model (Fig. 7). However, there is evidence that this may be too simplistic, for example, the type IIa RPTPs have also been reported in postsynaptic compartments7. The extent of exibility we observe for the RPTPs ectodomain, suggests that the RPTPs:TrkC binding mode revealed by our crystal structures may also mediate cis interactions in the event of co-localization of the two receptors at the same cell surface.

Our data reveal several key properties that t RPTPs for its dual role as a synaptic signalling hub and a promoter of neuronal growth. During axonal extension, RPTPs interacts with proteoglycan molecules through its N-terminal Ig1 domain, the clustering properties of HSPGs promoting growth cone motility14. The RPTPs ectodomain architecture described here

indicates that a range of conformations can be explored, permissive of cis interactions at the axonal surface and trans interactions to the general extracellular milieu including the basement membrane (Fig. 7a). Indeed, the length of the RPTPs ectodomain may be important to extend the HSPG-binding site beyond a saturating layer of cis interactions at the same cell surface, similar to the sialic acid-binding Siglec family of cell surface receptors where a lengthy ectodomain is required to escape the inhibitory glycocalyx40. At the transition from extension to synaptogenesis, the postsynaptic neuronal surface presents an array of additional RPTPs ligands (Fig. 7b)8,9.

Synapse formation and development requires the selection of an appropriate subset of binding partners. This involves a simple kinetic competition for binding, governed by inter-molecular afnity and interaction site accessibility. Our analyses suggest that

8 NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209 ARTICLE

this selection may be choreographed by the interplay of binding site location and conformational exibility in the RPTPs ectodomain. Structural comparison of the RPTPs:TrkC complex with our previously reported interaction mode for type IIa RPTP:GAGs14 shows partially overlapping binding sites (Fig. 6d). Thus, during synapse formation, and the shift from axonal growth to synaptic stability, postsynaptic TrkC must out-compete proteoglycans for RPTPs binding, simultaneously providing an adhesive trans interaction and extinguishing the RPTPs-clustering activity of HSPGs (Fig. 7b). Neuronal and glia-released proteoglycans continue to play important roles at synapses41,42, however, their contribution may involve other trans-synaptic interactions, for instance other type IIa RPTP family members or LRRTM4 (refs 43,44).

Alternative splicing has been shown previously to control the specicity of trans-synaptic interactions of the type IIa RPTPs8,17,20,28,29. The impact of alternative splicing at RPTPs mini-exon site meB on TrkC binding afnity provides a potential rheostat by which to ne-tune the balance between competing RPTPs ligands, an analogous feature to neurexin splice site 4 control of ligand interactions at the synaptic cleft8,45,46. The meA mini-exon is absent from all the RPTPs constructs used in this study, but its insertion into the Ig2 bDbE loop (b11b12 loop in Supplementary Fig. 1) is unlikely to affect binding of RPTPs to TrkC, as it would lie on the opposite face to the TrkC-binding interface. Both meA and meB sites are remote from the GAG-binding surface on RPTPs Ig1 (ref. 14) and therefore neither insertion is expected to modulate binding of RPTPs to proteoglycans.

Trans-synaptic binding to TrkC will limit the conformational freedom of the RPTPs ectodomain. This reduces the entropic

penalty for binding at other sites on RPTPs, an effect which can potentiate the formation of cell surface assemblies involving multiple receptor interactions47. For RPTPs function at the synapse this may facilitate cooperative binding of TrkC and NGL-3, as these two binding sites are separate18 (Fig. 7b). TrkC can also bind the NT-3 neurotrophin, a modulator of synaptic transmission, at the second Ig domain48 (Fig. 7b), which adds a further, distinctive stoichiometry to this network of trans-synaptic interactions, by specically triggering dimerization of TrkC and hence RPTPs. Other soluble modulators, such as the astrocyte-derived HSPGs glypican-4 and glypican-6, may provide an alternative strategy to regulate this system41. The stoichiometry and architecture of higher-order trans-synaptic complexes, and their impact on RPTPs enzymatic activity, remain important questions to address in the future. Formally, we cannot yet exclude a scenario where the simple kinetics of competition between ligands may be sufcient to explain the phenotypes observed in our cellular assays. However, given the geometrical constraints within which the RPTPs ectodomain has to operate at points of cellular contact, with ligands present in both cis and trans orientations, it is very likely that the RPTPs ectodomain exibility is required in a physiological context.

A series of proteolytic processing events have been reported for the type IIa RPTPs at the cell surface, involving initial shedding of the ectodomain, prior to release of the intracellular catalytic domains49. In our crystallization trials, we have potentially identied a further receptor-cleavage site at the consensus furin-like protease motif RVRR on the Ig2Ig3 linker, which would be disrupted in RPTPs isoforms containing mini-exon meB.

Cleavage of either RPTPs Ig12 or ectodomain fragments would rst decouple the receptor phosphatase activity from regulation via both proteoglycan and TrkC binding, and second the released soluble RPTPs fragments would be able to compete with the remaining intact receptors for binding to these same ligands. Knockout of the Drosophila type IIa RPTP dLAR, can be rescued by reintroduction of a catalytically inactive receptor, but not by a dLAR construct lacking the second inactive phosphatase domain (D2)50. It remains to be determined whether RPTPs ectodomainligand interactions may regulate the availability of RPTPs D2 for binding downstream intracellular partners8,9, but any such mode of regulation would also be ablated by extracellular RPTPs cleavage events.

In conclusion, our results suggest how the RPTPs nexus utilizes a series of ltering mechanisms to discriminate between binding options, and ultimately integrate the signalling inputs essential for the transition from neuronal growth to synapse organization. Rigidity has been demonstrated to be central for the comparatively simpler adhesion molecule function2426. In contrast, the properties of RPTPs described here, prompt the notion of how ectodomain exibility can allow a cell surface receptor to integrate a broad spectrum of ligand interactions into distinct functional outcomes.

Methods

Construct design and cloning. Human RPTPs Ig13, Ig1-FN3 and sEcto pHLsec constructs were reported previously14. The chick RPTPs Ig13 construct (amino acids 29316, NCBI Ref. Seq. NM_205407.1) was cloned into pHLsec and pHLAvitag3 vectors51. A series of chick RPTPs Ig13 mutant constructs were generated by PCR: K67A K68A K70A K71A (chick RPTPs Ig13 DK),

N73S S74N, R96A R99A, P97V T98H, K203A, Y223S and R227A R228A

and subsequently cloned into the pHL-Avitag3 vector. Human RPTPs Ig13 K68A K69A K71A K72A (human RPTPs Ig13 DK), R97A R100A,

Y224S and R228A R229A C-terminal Avitag constructs were similarly

constructed. Human RPTPd sEcto (amino acids 21833, NCBI Ref. Seq. BC106713.1), and human RPTP LAR Ecto (amino acids 301,260, NCBI Ref. Seq. NM_002840.3) were also cloned into the pHL-Avitag3 vector.

A synthetic clone for chick TrkC was commercially synthesized (Source

Bioscience); to include amino-acid residues 32302 (NCBI Ref. Seq.

HS

Growth cone

RPTP

HSPGs

TrkC NGL-3

NT3

Postsynapse

Presynapse

Synapse

Synaptic cleft

Figure 7 | Model illustrating exible RPTPr ectodomain sampling of extracellular ligands. (a) At the growth cone, RPTPs interacts with cell surface and basal membrane proteoglycans to mediate axonal extension.

(b) Upon contact with target cells, to shift to the role of synaptic organizer, RPTPs adopts elongated conformations to protrude from the presynaptic proteoglycan haze and bind postsynaptic ligands such as TrkC and NGL-3.

Subsequent independent or coordinated interactions with additional synaptic ligands are shown. Red boxes (left hand panels) indicate growth cone (a) or synapse (b) regions that are illustrated in the right-hand cartoons.

NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209

NM_205169.1), but lack residues 6377 and include the following point mutations: N163Q, N232Q, N259Q, N267Q and N294Q, to reduce the number of N-linked glycosylation sites. This construct (chick TrkC LRRIg1cryst) was cloned into the

pHL-Avitag3 vector. A series of further chick TrkC LRRIg1 mutant constructs were generated by PCR: D240A D242A, re-addition of residues 6377 (chick TrkC

LRRIg12N) and re-addition of residues 6377 with N68Q N72Q point mutations

(chick TrkC LRRIg12Q).

A series of mouse TrkC constructs (NCBI Ref. Seq. BC139764.1) were cloned into both pHLsec and pHL-Avitag3 vectors: LRR (amino acids 32208), LRRIg1 (amino acids 32302) and LRRIg2 (amino acids 32398). A mouse TrkC TM (amino acids 32463) construct was cloned into the pHLsec-monoVenus vector. The following mouse TrkC mutant constructs were generated by PCR:D240A D242A (LRRIg1 and TM constructs), N68Q N72Q (LRRIg12Q and

TM2Q) and removal of residues 6377 (LRRIg1cryst). Mouse NGL-3 (NCBI Ref. Seq. BC060263.1) ectodomain (NGL-3 Ecto; amino acids 1574) and full-length (NGL-3 FL; amino acids 1709) constructs were cloned into pHL-Avitag3 and pHLsec-monoCerulean vectors, respectively.

Protein purication and crystallization. For crystallization purposes, constructs were expressed in either HEK-293S GnTI cells (chicken RPTPs Ig13, TrkC

LRRIg1cryst and TrkC LRRIg12Q) or HEK293-T cells treated with kifunensine

(human RPTPs Ig1-FN3) following transient transfection using polyethylenimine51. The proteins were puried from 0.2-mm-ltered cell culture media by immobilized nickel afnity chromatography (Chelating Sepharose Fast Flow, GE Healthcare) followed by size-exclusion chromatography in 10 mM HEPES, 150 mM NaCl, pH 7.5.

Crystallization trials, using 100 nl protein solution plus 100 nl reservoir solution in sitting-drop vapour diffusion format were set up in 96-well Greiner plates using a Cartesian Technologies robot, and plates were subsequently maintained at20.5 C in a TAP Homebase storage vault. The crystallization conditions yielding diffraction quality crystals were: 10% polyethylene glycol (PEG) 20 k, 20% PEG MME 550, 0.1 M bicine/Tris pH 8.5, 0.03 M sodium nitrate, 0.03 M disodium hydrogen phosphate, 0.03 M ammonium sulphate (chicken RPTPs Ig12:TrkC

LRRIg1cryst complex 2.5 data set), 10% w/v PEG MME 5 k, 0.1 M HEPES pH 7.0, 5% w/v Tacsimate (chicken RPTPs Ig13:TrkC LRRIg12Q complex 3.05 data set) and 10% PEG 400, 0.01 M magnesium chloride, 0.1 M potassium chloride, 0.05 M MES pH 6.0 (human RPTPs Ig1-FN3).

Data collection and processing. Crystals were cryoprotected using a 2530% solution of ethylene glycol (chicken RPTPs Ig12:TrkC LRRIg1cryst and RPTPs

Ig13:TrkC LRRIg12Q complexes) or 25% propylene glycol (human RPTPs Ig1-FN3) and then ash-cooled at 100 K. X-ray diffraction data were collected at the I03 (chicken RPTPs Ig13:TrkC LRRIg12Q; wavelength 0.9763 ) and I04 (chicken

RPTPs Ig12:TrkC LRRIg1cryst; wavelength 0.9200 ) beamlines Diamond Light

Source, Oxfordshire, UK and the ID29 (human RPTPs Ig1-FN3; wavelength0.9788 ) beamline at the European Synchotron Radiation Facility, Grenoble, France. The diffraction images were indexed, integrated, scaled and merged using the xia2 data-processing suite52. We used Rpim, I/sI and CC1/2 (ref. 53) statistics in the highest-resolution shell (with criteria Rpimo100%, I/sI41.5 and CC1/2450%)

to determine our high-resolution cutoffs (see Table 1).

Molecular replacement was used to phase all three crystal structures, using human and chicken RPTPs Ig12 (PDB ID: 2YD3 and 2YD4), human RPTPs Ig3 (from PDB ID: 2YD9), a Ca model for human RPTP LAR FN4 (PDB ID: 2DJU) and models of the chicken TrkC LRR and rst Ig domain (Ig1) generated using the SWISS-MODEL web interface54, as search models in Phaser55. Manual model adjustment was performed in Coot56 and the Refmac57, Phenix58 and Buster (Global Phasing Ltd)59 suites used for renement (applying translation libration screw-motion restraints for all structures and local non-crystallographic symmetry restraints for RPTPs:TrkC complex structures). Stereochemical properties of all models were accessed using MolProbity60. Ramachandran statistics: human RPTPs Ig1-FN3, 94.0% most favoured, 6.0% additionally allowed and no disallowed rotamers; chicken RPTPs Ig12:TrkC LRRIg1cryst, 98% most favoured,

2% additionally allowed and no disallowed rotamers; chicken RPTPs Ig13:TrkC LRRIg12Q, 96.6% most favoured, 3.2% additionally allowed and 0.2% disallowed rotamers. Full data collection and renement statistics are given in Table 1.

The 3.15- human RPTPs Ig1-FN3 crystal structure contains one molecule per asymmetric unit; amino residues 35601, an additional C-terminal G residue derived from the expression vector and two GlcNAc residues at N-linked glycosylation sites N250 and N295.

The 2.5 chicken RPTPs Ig12:TrkC LRRIg1cryst crystal structure contains amino-acid residues 32302 for two molecules, A and B, of TrkC LRRIg1 (lacking residues 6377 absent in the TrkC LRRIg1cryst expression construct and residues 258261 on a disordered Ig1 loop, but with ETG (A) and G (B) additional N-term residues and GT (A) and G (B) additional C-term residues derived from the expression plasmid) and residues 29227, 29226 and 29228 for three molecules, C, D and E, of RPTPs Ig12 (with the exception of a disordered loop for each molecule, 6871(C), 6870(D) and 6873(E)). Five TrkC N-linked glycosylation sites were modelled; initial N-acetylglucosamine (GlcNAc) residues, covalently bonded to N79 (A and B), N203 (A) and N272 (A and B) were included in the structure. Inspection of the crystal packing clearly demonstrates that the Ig3

domains of the RPTPs Ig13 crystallization protein are not present in these crystals. Proteolytic cleavage is the most likely reason for their absence, as we have commonly seen cleavage of wild-type RPTPs Ig13 at the Ig2Ig3 linker during crystallization trials. The previously reported human RPTPs Ig12 crystal structure was obtained via proteolysis of Ig13 during crystallization (PDB ID 2YD3). To produce crystals of intact human RPTPs Ig13, R227Q R228N

(residue numbering relative to chicken RPTPs) point mutations were previously introduced to disrupt a potential furin protease motif14. However, RPTPs R227 and R228 are required for full RPTPs:TrkC binding (Fig. 4c; SupplementaryFig. 8e,f; Supplementary Table 1) and therefore the RPTPs Ig13 R227Q R228N

protein was not used in crystallization trials in this study.

The 3.05 chicken RPTPs Ig13:TrkC LRRIg12Q crystal structure contains amino-acid residues 32302 for three molecules, A, B and C, of TrkC LRRIg1 (lacking residues 5968 (A), 5968 (B) and 5773 (C), but with TG (A), G (B) and G (C) additional N-term residues derived from the expression plasmid) and residues 29226, 29227 and 29225 for three molecules, D, E and F, of RPTPs

Ig13 (with the exception of residues 6771 in the disordered Lys-loop for each molecule, for which density was visible, but a single conformation could not be built). Electron density was visible to indicate the presence of N-linked glycosylation adjacent to N79 in TrkC molecules AC, but only in molecule A could an initial GlcNAc residue be successfully rened. There was also evidence that N218 (electron density corresponding to molecule B) and N272 (electron density corresponding to molecule A) are also N-linked glycosylation sites. Sparse electron density is visible for RPTPs Ig3, suggesting disorder/multiple conformations of this domain.

The assignment of secondary-structure elements was performed using ksdssp61. The superposition of atomic models to compare the domain architecture between different structures was performed using SHP62, based on Ca positions.

Crystallographic gures were created using PyMOL (Schrdinger, LLC) and APBS63 was used to calculate the electrostatic potential of solvent-accessible surfaces.

Surface Plasmon Resonance. All SPR experiments were performed on BIAcore T100 or T200 instruments. Ligand constructs were expressed in HEK293-T cells with no kifunensine treatment and puried via Ni-afnity chromatography. All contained a C-terminal BirA recognition site (Avitag) and were biotinylated enzymatically prior to immobilization to the surface of CM5 sensor chips (BIA-core) pre-coated with streptavidin (Sigma) using the BIAcore amine-coupling kit. Analyte constructs were expressed in either GnTI HEK293-S or HEK293-T cells and puried as for crystallization purposes, described above.

Unless otherwise stated, experiments were all performed in 10 mM HEPES pH7.5, 150 mM NaCl, 0.01% Tween 20, at a temperature of 25 C and a owrate of 20 ml min 1. Typically, each analyte sample was injected across the chip surfaces for 120 s and then a 300 s dissociation time was included to allow the signal to return to baseline. For equilibrium SPR experiments, serial two- or threefold dilutions of protein analyte were sequentially injected and all injection series were repeated in duplicate. No regeneration of the chip surfaces was required between analyte injections, except when measuring the interaction between chicken TrkC LRRIg12Q and chicken RPTPs Ig13, when a 30 s injection of 1 M MgCl2 was sufcient for the signal to return to baseline.

Scrubber2 (BioLogic Software) and Prism (GraphPad Software) were used for data analysis assuming the Langmuir model and a 1:1 ligand to analyte ratio. The signal for experimental ow cells was corrected by initial subtraction of a blank (only buffer injected as analyte across the ow cell of interest), followed by the subtraction of the reference signal from a mock-coupled ow cell (streptavidin, but no ligand bound). To estimate half-maximal inhibitory concentration values for heparan sulphate or heparin-dp10 inhibition of RPTPs:TrkC binding, nonlinear regression in Prism (GraphPad Software) was used to t a variable-slope dose response curve to the experimental data, xing top and bottom values according to control measurements (top; response units when no inhibitor present and bottom; response units when no protein analyte or inhibitor present), while the Hill Slope coefcient and half-maximal inhibitory concentration variables were left unrestrained.

Multi-Angle Light Scattering. MALS experiments were carried out on a Wyatt MALS/AFFFF System (Wyatt Technologies). Human RPTPs Ig1-FN3 and mouse TrkC LRRIg1 proteins were expressed in GnTI HEK-293S cells and puried as described above. Size-exclusion chromatography was performed in 10 mM Tris, 50 mM NaCl, pH 7.5 on a Superdex200 HR10/30 column (GE Healthcare), attached to an Agilent chromatography system. An Optilab rEX Refractive Index detector and a Dawn Helios II Multi-Angle Light Scattering detector recorded the refractive index and light scattering of the samples upon elution from the size-exclusion column. The Wyatt software ASTRA was used to analyse all the data collected.

Negative-stain EM. Human RPTPs Ig1-FN3 and sEcto proteins were negatively stained with 0.7% uranyl formate64. Images were recorded using an FEI electron microscope equipped with a LaB6 lament operated at an acceleration voltage of 200 keV at a magnication of 55,000 and a defocus value of approximately

10 NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209 ARTICLE

1.5 mm. All images were recorded using SO-163 lm and developed with a Kodak D-19 developer at full strength for 12 min at 20 C. The electron micrographs were digitized with a CoolScan 9000 (Nikon) using a step size6.35 mm, and the pixels were binned by a factor of 3. As a result, the specimen level pixel size was at 3.8 . To generate projection averages, particles were interactively selected using the WEB display program in SPIDER65 and windowed into 90 90-

pixel (human RPTPs Ig1-FN3) and 100 100-pixel (human RPTPs sEcto)

images. Class averages were calculated using these windowed images over 10 cycles of K-means classication and multi-reference alignment specifying 150 classes64.

SAXS. RPTPs Ig1-FN3 and sEcto proteins were deglycosylated by Endo-F1 and puried by SEC immediately prior to data collection. Solution scattering data were collected at beamline BM29 of the European Synchotron Radiation Facility66 at 293 K within a momentum transfer range of 0.01 1oqo0.45 1, whereq 4psin(y)/l and 2y is the scattering angle. X-ray wavelength was 0.995 and

data were collected on a Pilatus 1M detector. RPTPs Ig1-FN3 was measured at1.33 and 5.33 g l 1 and RPTPs sEcto at 1.00 and 4.00 g l 1. Data reduction and calculation of invariants was carried out using standard protocols implemented in the ATSAS software suite67. A merged data set was obtained by merging the low-angle part of the low-concentration data set with the high-angle part of the high-concentration data set.

A pool of 10,000 independent model conformers was constructed using the program RANCH67 for both RPTPs Ig1-FN3 and sEcto by treating individual domains as beads-on-a-string. For RPTPs sEcto, a homology model for domains 8 and 9 was created using SWISS-MODEL54. Both pools contained conformer shapes ranging from collapsed or U-shaped to fully extended, as evidenced by their Gaussian RG and DMAX distributions. Ensemble selection using the experimental

SAXS data as constraint with the programs GAJOE67 or MES68 indicated predominantly extended conformers, consistent with the experimentally determined RG and DMAX. The MES-selected models were used as starting structures for further modelling. Missing loops and N- and C termini were added in extended conformations using the program Modeller. All-atom simulations of RPTPs Ig1-FN3 and sEcto was performed using the program AllosMod. For each starting structure, 30 independent pools of 100 models were generated. For the combined pool, calculation and tting of scattering patterns was performed using the program FoXS, and automated selection of the minimal set of models satisfying the scattering data was performed using the program MES; this whole procedure was automated using the AllosMod-FoXS web server68.

Neuron-COS cell co-culture assays. Dissociated primary hippocampal neuron cultures were prepared from embryonic day 18 rat embryos69. For co-culture assays, COS-7 cells were transfected and 24 h later were seeded onto neurons at 14 days in vitro70. As indicated, co-culture coverslips were incubated with 0.2 U ml 1 heparinase I, II and III or with 30 mg ml 1 (B10 mM) heparin-dp10 (Iduron Ltd,

UK) in glial conditioned medium. Co-cultures were xed with 4% formaldehyde and 4% sucrose in phosphate-buffered saline (PBS) (pH 7.4) followed by permeabilization with 0.2% Triton X-100. For cell surface staining of TrkC, COS-7 cells were xed without permeabilization. Fixed cultures were blocked in 3% bovine serum albumin and 5% normal goat serum in PBS for 30 min at 37 C, and primary antibodies (overnight incubation at 4 C) then secondary antibodies (1 h at 37 C) were applied in 3% bovine serum albumin and 5% normal goat serum in PBS. Coverslips were mounted in elvanol (Tris-HCl, glycerol and polyvinyl alcohol, with 2% 1,4-diazabicyclo[2,2,2]octane). The primary antibodies were: anti-TrkC (1:500; C44H5; Cell Signaling), anti-synapsin I (rabbit, 1:2,000; Millipore; AB1543P) for presynaptic terminals, anti-MAP2 (chicken polyclonal IgY; 1:2,000; Abcam; ab5392) for dendrites and anti-dephospho-tau (mouse mIgG2a 1:2,000, clone PC1C6; Millipore; MAB3420) for axons.

All image acquisitions, analyses and quantications were performed by investigators blind to the experimental condition. For co-cultures, elds for imaging were chosen only by the Cyan Fluorescent Protein (CFP) or mVenus and phase-contrast channels, for the presence of CFP or mVenus-positive COS-7 cells in a neurite-rich region. The synapsin channel was thresholded and the total intensity of puncta within all regions positive for both CFP or mVenus and dephospho-tau but negative for MAP2 was measured. Analysis was performed using Fiji (ImageJ2), and GraphPad Prism 5. All data are reported as means.e.m.

References

1. Siddiqui, T. J. & Craig, A. M. Synaptic organizing complexes. Curr. Opin. Neurobiol. 21, 132143 (2011).

2. Shen, K. & Scheiffele, P. Genetics and cell biology of building specic synaptic connectivity. Annu. Rev. Neurosci. 33, 473507 (2010).

3. Good, M. C., Zalatan, J. G. & Lim, W. A. Scaffold proteins: hubs for controlling the ow of cellular information. Science 332, 680686 (2011).

4. Craig, A. M. & Kang, Y. Neurexin-neuroligin signaling in synapse development. Curr. Opin. Neurobiol. 17, 4352 (2007).

5. Bell, C. H. et al. Structure of the repulsive guidance molecule (RGM)-neogenin signaling hub. Science 341, 7780 (2013).

6. Johnson, K. G. & Van Vactor, D. Receptor protein tyrosine phosphatases in nervous system development. Physiol. Rev. 83, 124 (2003).

7. Dunah, A. W. et al. LAR receptor protein tyrosine phosphatases in the development and maintenance of excitatory synapses. Nat. Neurosci. 8, 458467 (2005).

8. Takahashi, H. & Craig, A. Protein tyrosine phosphatases PTPd, PTPs, and LAR: presynaptic hubs for synapse organization. Trends Neurosci. 36, 522534 (2013).

9. Um, J. W. & Ko, J. LAR-RPTPs: synaptic adhesion molecules that shape synapse development. Trends Cell Biol. 23, 465475 (2013).

10. Aricescu, A. R., McKinnell, I. W., Halfter, W. & Stoker, A. W. Heparan sulfate proteoglycans are ligands for receptor protein tyrosine phosphatase s. Mol.

Cell. Biol. 22, 18811892 (2002).11. Fox, A. N. & Zinn, K. The heparan sulfate proteoglycan syndecan is an in vivo ligand for the drosophila LAR receptor tyrosine phosphatase. Curr. Biol. 15, 17011711 (2005).

12. Johnson, K. G. et al. The HSPGs syndecan and dallylike bind the receptor phosphatase LAR and exert distinct effects on synaptic development. Neuron 49, 517531 (2006).

13. Shen, Y. et al. PTPsigma is a receptor for chondroitin sulfate proteoglycan, an inhibitor of neural regeneration. Science 326, 592596 (2009).

14. Coles, C. H. et al. Proteoglycan-specic molecular switch for RPTPs clustering and neuronal extension. Science 332, 484488 (2011).

15. Fisher, D. et al. Leukocyte common antigen-related phosphatase is a functional receptor for chondroitin sulfate proteoglycan axon growth inhibitors.J. Neurosci. 31, 1405114066 (2011).16. Fry, E. J., Chagnon, M. J., Lpez-Vales, R., Tremblay, M. L. & David, S. Corticospinal tract regeneration after spinal cord injury in receptor protein tyrosine phosphatase sigma decient mice. Glia 58, 423433 (2010).

17. Takahashi, H. et al. Postsynaptic TrkC and presynaptic PTPs function as a bidirectional excitatory synaptic organizing complex. Neuron 69, 287303 (2011).

18. Kwon, S. K., Woo, J., Kim, S. Y., Kim, H. & Kim, E. Trans-synaptic adhesions between netrin-G ligand-3 (NGL-3) and receptor tyrosine phosphatasesLAR, protein-tyrosine phosphatase delta (PTPdelta), and PTPsigma via specic domains regulate excitatory synapse formation. J. Biol. Chem. 285, 1396613978 (2010).

19. Yoshida, T. et al. Interleukin-1 receptor accessory protein organizes neuronal synaptogenesis as a cell adhesion molecule. J. Neurosci. 32, 25882600 (2012).

20. Yim, Y. S. et al. Slitrks control excitatory and inhibitory synapse formation with LAR receptor protein tyrosine phosphatases. Proc. Natl Acad. Sci. USA 110, 40574062 (2013).

21. Pulido, R., Serra-Pags, C., Tang, M. & Streuli, M. The LAR/PTP delta/PTP sigma subfamily of transmembrane protein-tyrosine-phosphatases: multiple human LAR, PTP delta, and PTP sigma isoforms are expressed in a tissue-specic manner and associate with the LAR-interacting protein LIP.1. Proc. Natl Acad. Sci. USA 92, 1168611690 (1995).

22. Biersmith, B. H., Hammel, M., Geisbrecht, E. R. & Bouyain, S. The immunoglobulin-like domains 1 and 2 of the protein tyrosine phosphatase LAR adopt an unusual horseshoe-like conformation. J. Mol. Biol. 408, 616627 (2011).

23. Comoletti, D. et al. The macromolecular architecture of extracellular domain of aNRXN1: domain organization, exibility, and insights into trans-synaptic disposition. Structure 18, 10441053 (2010).

24. Brasch, J., Harrison, O. J., Honig, B. & Shapiro, L. Thinking outside the cell: how cadherins drive adhesion. Trends Cell Biol. 22, 299310 (2012).

25. Aricescu, A. R. et al. Structure of a tyrosine phosphatase adhesive interaction reveals a spacer-clamp mechanism. Science 317, 12171220 (2007).

26. Ozkan, E. et al. Extracellular architecture of the SYG-1/SYG-2 adhesion complex instructs synaptogenesis. Cell 156, 482494 (2014).

27. Woo, J. et al. Trans-synaptic adhesion between NGL-3 and LAR regulates the formation of excitatory synapses. Nat. Neurosci. 12, 428437 (2009).

28. Takahashi, H. et al. Selective control of inhibitory synapse development by Slitrk3-PTPd trans-synaptic interaction. Nat. Neurosci. 15, 389398 (2012).

29. Yoshida, T. et al. IL-1 receptor accessory protein-like 1 associated with mental retardation and autism mediates synapse formation by trans-synaptic interaction with protein tyrosine phosphatase d. J. Neurosci. 31, 1348513499 (2011).

30. Valnegri, P. et al. The X-linked intellectual disability protein IL1RAPL1 regulates excitatory synapse formation by binding PTPd and RhoGAP2. Hum.

Mol. Genet. 20, 47974809 (2011).31. Wang, H. et al. Expression of receptor protein tyrosine phosphatase-s (RPTP-s) in the nervous system of the developing and adult rat. J. Neurosci. Res. 41, 297310 (1995).

32. Valenzuela, D. M. et al. Alternative forms of rat TrkC with different functional capabilities. Neuron 10, 963974 (1993).

33. Lein, E. S. et al. Genome-wide atlas of gene expression in the adult mouse brain. Nature 445, 168176 (2007).

34. Elchebly, M. et al. Neuroendocrine dysplasia in mice lacking protein tyrosine phosphatase s. Nat. Genet. 21, 330333 (1999).

NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6209

35. Wallace, M. J. et al. Neuronal defects and posterior pituitary hypoplasia in mice lacking the receptor tyrosine phosphatase PTPs. Nat. Genet. 21, 334338 (1999).

36. Tessarollo, L. et al. Targeted deletion of all isoforms of the trkC gene suggests the use of alternate receptors by its ligand neurotrophin-3 in neuronal development and implicates trkC in normal cardiogenesis. Proc. Natl Acad. Sci. USA 94, 1477614781 (1997).

37. Klein, R. et al. Disruption of the neurotrophin-3 receptor gene trkC eliminates la muscle afferents and results in abnormal movements. Nature 368, 249251 (1994).

38. Horn, K. E. et al. Receptor protein tyrosine phosphatase sigma regulates synapse structure, function and plasticity. J. Neurochem. 122, 147161 (2012).

39. Chen, X., Yoshida, T., Sagara, H., Mikami, Y. & Mishina, M. Protein tyrosine phosphatase sigma regulates the synapse number of zebrash olfactory sensory neurons. J. Neurochem. 119, 532543 (2011).

40. Razi, N. & Varki, A. Masking and unmasking of the sialic acid-binding lectin activity of CD22 (Siglec-2) on B lymphocytes. Proc. Natl Acad. Sci. USA 95, 74697474 (1998).

41. Allen, N. et al. Astrocyte glypicans 4 and 6 promote formation of excitatory synapses via GluA1 AMPA receptors. Nature 486, 410414 (2012).

42. Irie, F., Badie-Mahdavi, H. & Yamaguchi, Y. Autism-like socio-communicative decits and stereotypies in mice lacking heparan sulfate. Proc. Natl Acad. Sci. USA 109, 50525056 (2012).

43. Siddiqui, T. J. et al. An LRRTM4-HSPG complex mediates excitatory synapse development on dentate gyrus granule cells. Neuron 79, 680695 (2013).44. de Wit, J. et al. Unbiased discovery of glypican as a receptor for LRRTM4 in regulating excitatory synapse development. Neuron 79, 696711 (2013).

45. Uemura, T. et al. Trans-synaptic interaction of GluRdelta2 and neurexin through Cbln1 mediates synapse formation in the cerebellum. Cell 141, 10681079 (2010).

46. Siddiqui, T. J., Pancaroglu, R., Kang, Y., Rooyakkers, A. & Craig, A. M. LRRTMs and neuroligins bind neurexins with a differential code to cooperate in glutamate synapse development. J. Neurosci. 30, 74957506 (2010).

47. Wu, Y., Vendome, J., Shapiro, L., Ben-Shaul, A. & Honig, B. Transforming binding afnities from three dimensions to two with application to cadherin clustering. Nature 475, 510513 (2011).

48. Park, H. & Poo, M. M. Neurotrophin regulation of neural circuit development and function. Nat. Rev. Neurosci. 14, 723 (2013).

49. Aicher, B., Lerch, M. M., Muller, T., Schilling, J. & Ullrich, A. Cellular redistribution of protein tyrosine phosphatases LAR and PTPsigma by inducible proteolytic processing. J. Cell Biol. 138, 681696 (1997).

50. Krueger, N. X. et al. Functions of the ectodomain and cytoplasmic tyrosine phosphatase domains of receptor protein tyrosine phosphatase Dlar in vivo. Mol. Cell. Biol. 23, 69096921 (2003).

51. Aricescu, A. R., Lu, W. & Jones, E. Y. A time- and cost-efcient system for high-level protein production in mammalian cells. Acta Crystallogr. D Biol. Crystallogr. 62, 12431250 (2006).

52. Winter, G. xia2: an expert system for macromolecular crystallography data reduction. J. Appl. Crystallogr. 43, 186190 (2010).

53. Karplus, P. A. & Diederichs, K. Linking crystallographic model and data quality. Science 336, 10301033 (2012).

54. Arnold, K., Bordoli, L., Kopp, J. & Schwede, T. The SWISS-MODEL workspace: a web-based environment for protein structure homology modelling. Bioinformatics 22, 195201 (2006).

55. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40,

658674 (2007).

56. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 21262132 (2004).

57. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. Renement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240255 (1997).

58. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213221 (2010).

59. Blanc, E. et al. Renement of severely incomplete structures with maximum likelihood in BUSTER-TNT. Acta Crystallogr. D Biol. Crystallogr. 60, 22102221 (2004).

60. Davis, I. W. et al. MolProbity: all-atom contacts and structure validation for proteins and nucleic acids. Nucleic Acids Res. 35, W375W383 (2007).

61. Kabsch, W. & Sander, C. Dictionary of protein secondary structure: Pattern recognition of hydrogen-bonded and geometrical features. Biopolymers 22, 25772637 (1983).

62. Stuart, D. I., Levine, M., Muirhead, H. & Stammers, D. K. Crystal structure of cat muscle pyruvate kinase at a resolution of 2.6 . J. Mol. Biol. 134, 109142 (1979).

63. Baker, N. A., Sept, D., Joseph, S., Holst, M. J. & McCammon, J. A. Electrostatics of nanosystems: Application to microtubules and the ribosome. Proc. Natl Acad. Sci. USA 98, 1003710041 (2001).

64. Ohi, M., Li, Y., Cheng, Y. & Walz, T. Negative staining and image classicationpowerful tools in modern electron microscopy. Biol. Proced. Online 6, 2334 (2004).

65. Frank, J. et al. SPIDER and WEB: processing and visualization of images in 3D electron microscopy and related elds. J. Struct. Biol. 116, 190199 (1996).

66. Pernot, P. et al. Upgraded ESRF BM29 beamline for SAXS on macromolecules in solution. J. Synchrotron. Radiat. 20, 660664 (2013).

67. Petoukhov, M. V. et al. New developments in the ATSAS program package for small-angle scattering data analysis. J. Appl. Crystallogr. 45, 342350 (2012).

68. Guttman, M., Weinkam, P., Sali, A. & Lee, K. K. All-atom ensemble modeling to analyze small-angle X-ray scattering of glycosylated proteins. Structure 21, 321331 (2013).

69. Kaech, S. & Banker, G. Culturing hippocampal neurons. Nat. Protoc. 1, 24062415 (2006).

70. Graf, E. R., Zhang, X., Jin, S.-X., Linhoff, M. W. & Craig, A. M. Neurexins Induce differentiation of GABA and glutamate postsynaptic specializations via neuroligins. Cell 119, 10131026 (2004).

Acknowledgements

Coordinates and structure factors are deposited in the Protein Data Bank with PDB IDs: 4PBV, 4PBW and 4PBX. We thank K. Harlos, T. Walter, A. Clayton, G. Sutton,N. Shanks and D. Staunton for assistance, F. Alder and L. Ciani (UCL) for provision of hippocampal cells in preliminary studies. We also thank staff at the European Synchrotron Radiation Facility (beamlines ID29 and BM29) and Diamond Light Source (beamlines I03 and I04, proposal mx8423). This work was funded by the UK Medical Research Council (G0700232 and L009609 to A.R.A. and G9900061 to E.Y.J.), Cancer Research UK (A10976 to E.Y.J.) and USA National Institutes of Health (MH070860 to A.M.C. and R01HD061543 from NICHD to T.N.). The Wellcome Trust Centre for Human Genetics is supported by Wellcome Trust grant 090532/Z/09/Z. N.M. was the recipient of a Wellcome Trust D.Phil. studentship, P.Z. is the recipient of a Michael Smith Foundation for Health Research Postdoctoral Fellowship, C.H.C. is a Human Frontiers Science Program long-term Postdoctoral Fellow (LT000100/2013), J.E. is supported by EMBO and Marie-Curie long-term postdoctoral fellowships, A.M.C. is a Canada Research Chair and A.R.A. is a UK Medical Research Council Senior Fellow.

Author contributions

A.R.A., E.Y.J. and C.H.C. designed the project and all authors contributed to data analysis and preparation of the manuscript. W.L. transiently expressed all proteins. C.H.C. and N.M. cloned and puried all constructs and performed SPR and crystallographic analyses for the RPTPs:TrkC complexes. C.H.C. solved the RPTPs Ig1-FN3 crystal structure. T.N. and C.H.C. performed the negative-stain electron microscopy and processed and analysed these data. N.M. and J.E. collected and analysed all SAXS data. A.W.S. carried out cellular assays. P.Z. performed COS-hippocampal cell culture experiments and analysed these data together with A.M.C.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Coles, C. H. et al. Structural basis for extracellular cis and trans RPTPs signal competition in synaptogenesis. Nat. Commun. 5:5209 doi: 10.1038/ncomms6209 (2014).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

12 NATURE COMMUNICATIONS | 5:5209 | DOI: 10.1038/ncomms6209 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.