Structural Basis of Teneurin-Latrophilin Interaction in Repulsive Guidance of Migrating Neurons

1Teneurins are eukaryotic cell adhesion receptors that are thought to have evolved through a horizontal gene transfer event, where fusion of a bacterial toxin gene to a eukaryotic receptor resulted in a large type II transmembrane protein. They have been described in worms, insects, vertebrates, and single-celled choanoflagellates. The four mammalian Teneurin homologs, Ten1 to Ten4, (also referred to as Ten-m1 to Ten-m4 and Odz-1 to Odz-4) are found predominantly, but not exclusively, in the nervous system, where they display complementary expression patterns from early development through to adulthood. In humans, Teneurins are risk loci in bipolar disorder, and schizophrenia, and Ten3 mutations have been implicated in microphthalmia and general anosmia. The mammalian Teneurins are modular, ∼2,800-residue-long type II transmembrane proteins whose ∼250-kDa extracellular domain comprises at least 16 annotated domains (Figure 1A). We recently revealed the first X-ray crystallography and cryoelectron microscopy structures of an ∼200-kDa fraction of the Gallus gallus Ten2 and murine Ten3 extracellular domains. These revealed a conserved superfold of eight domains. A similar cryoelectron microscopy structure of human Ten2 confirmed that key features of this fold are conserved. The structures revealed three signature motifs of the Teneurin fold: (1) the spiraling beta-barrel tyrosine-aspartate repeat “YD shell” domain, (2) a specialized “fibronectin plug” domain that seals off the YD shell at the N-terminal side, and (3) a beta-propeller referred to as the NCL-1, HT2A, and Lin-41 (NHL) domain. These three elements form a superfold that is widespread in bacterial genomes, suggesting that they represent an evolutionarily ancient uncharacterized family of secreted bacterial proteins. The YD shell bears structural similarity to bacterial toxins of the TcB-TcC family. The regions upstream and downstream of the fibronectin (FN) plug, NHL, and YD shell are highly conserved in mammalian Teneurins. C-terminal of the YD shell is an ∼200-amino-acid linker that resides in the YD repeat shell and leads through the shell wall to form the antibiotic-binding-like (ABD) and Tox-GHH domains. The Tox-GHH harbors a colicin-like DNase fold that includes a Teneurin C-terminal associated peptide, TCAP. This neuropeptide is either cleaved from the full Teneurin protein or transcribed separately and is thought to modulate murine stress behavior. Upstream of the Teneurin core fold lies the extracellular transthyretin (TTR)-like domain, a cysteine-rich region that is not structurally annotated, and eight epidermal growth factor (EGF) domains, two of which form disulphide bridges and mediate Teneurin dimer formation, presumably “in cis” (between Teneurins on the same cell). An ∼180-amino-acid-long linker leads from the EGF domains to the transmembrane (TM) helix and intracellular domain (ICD).

2Pioneering studies in flies, which have two Teneurins, Ten-a and Ten-m, revealed roles in synaptic matching in the olfactory system and at neuromuscular junctions. In mice, Ten3 and Ten2 direct axonal wiring in the visual system, the hippocampus, and of thalamostriatal projections. The NHL domain of mammalian Teneurins harbors an alternatively spliced loop (SS) that determines homophilic binding of Teneurin in cell adhesion. Teneurins promote synapse development and trans-synaptic adhesion by also engaging in heterophilic interactions with Latrophilins. Teneurin ectodomains that are proteolytically released from the cell surface bind to Latrophilins and act as attractants in axon guidance. Latrophilins (Lphn1–Lphn3 in mammals) are adhesion G-protein-coupled receptors (GPCRs) that have received attention because of their role as a calcium-independent receptor of latrotoxin, their recently reported functions as mechanosensitive receptors, and their roles in neurodevelopment disorders such as attention deficit hyperactivity disorder (ADHD). Latrophilins also interact with FN leucine-rich repeat proteins (FLRTs) in synaptogenesis and engage in large supercomplexes that include Uncoordinated-5 (Unc5) receptors. Latrophilins contain a short C-terminal ICD, a seven-TM GPCR domain (7TM), the extracellular GPCR autoproteolysis-inducing and hormone domains (GAIN/Horm), a glycosylated linker region (∼100 residues), and N-terminal olfactomedin (Olf) and lectin (Lec) domains (Figure 1A). FLRTs are single-spanning type I TM proteins with an extracellular leucine-rich repeat (LRR) domain that is connected to a downstream FN-like domain via a glycosylated linker (Figure 1A). The LRR domain of FLRT binds the Latrophilin Olf domain, whereas Teneurins require the Latrophilin Lec domain for binding. The linker between the Lec and Olf domains undergoes alternative splicing and determines the affinity of Latrophilin binding to Teneurin. Recent work shows that FLRT3, Lphn2/3, and Ten2 work together as coincidence receptors in directing hippocampal synapse formation. How Teneurins and Latrophilins interact at the molecular level and whether Teneurin-Latrophilin interactions have functions in early neurodevelopment prior to wiring has remained unclear. Here we ask the following questions. What is the structural mechanism of the Teneurin-Latrophilin interaction? Is it compatible with the known FLRT-Latrophilin binding mechanism? What is the function of the synaptic proteins Teneurin, Latrophilin, and FLRT during early cortex development? With X-ray crystallography, we reveal the Latrophilin binding site on Teneurin at the lateral side of the YD shell domain. The Teneurin-Latrophilin binding mechanism we reveal is consistent with coincidence binding of FLRT to the Latrophilin Olf domain, and we suggest a ternary complex model using our previous structural data on the Latrophilin-FLRT interaction. We use the structural results to probe for receptor functions in early cortical development and reveal a repulsion-mediated mechanism of cell body guidance. These results expand the functional repertoire of these receptors beyond a role in synapse development.

3Previous studies have demonstrated that the Lec domain of Latrophilin is essential and sufficient for binding to Teneurin proteins. We produced murine Latrophilin 2 (Lphn2 Lec domain, residues 30–137) and chicken Teneurin 2 (residues 1,043–2,802) individually in HEK293 cells and mixed the purified proteins. Crystals of Ten2 and Lphn2 grew in sitting drops at 4°C. We collected X-ray diffraction data up to 3.6-Å resolution and solved the structure by molecular replacement (Figure 1B; PDB: 6SKE). Crystallographic details are summarized in Table S1. The data reveal the Lphn2 Lec binding site on the Teneurin YD shell, spanning the tiers of the spiraling YD shell domain (Figure 1B) and burying a total of ∼1,500 Å2 surface area. The Olf domain is thought to contribute to the binding between Teneurin and Latrophilin. We therefore also expressed murine Latrophilin 1 (Lphn1 Lec-Olf domains, residues 29–395) in HEK293 cells and produced complex crystals with Ten2 (residues 955–2802) in sitting drops at 18°C. We collected X-ray diffraction data up to 3.86-Å resolution and solved the structure by molecular replacement (Figures 1C and S1A; PDB: 6SKA). This second structure confirms the conserved Lec binding site on the Ten2 YD shell domain (Figures 1B, S1B, and S1C). Most of the buried Teneurin surface is contacted by the Lec domain, which is consistent with previous reports showing that Lec is essential and sufficient for the interaction. The total buried surface in this complex is ∼2,500 Å2. In both structures, the C-terminal Tox-GHH/TCAP domain lies ∼6 nm away from the Latrophilin binding site (Figures 1D and 1E). FLRT is known to bind Latrophilin Olf via the concave surface of its LRR domain. Superposition of the previously solved Lphn3-FLRT2 complex structure produces a model of how the three proteins likely interact in a ternary complex (Figures 1C and S1D). Sequence conservation in the binding interfaces of vertebrate Latrophilins and Teneurins (Figures S1E and S1F) suggests that the interaction has conserved functional importance that restrains sequence diversion in this area.

4We chose the minimal complex of Lphn2 Lec-Ten2 (Figure 1B) for a set of molecular dynamics (MD) simulations to assess the stability of the complex and to better define the positions of side chains in the interface. A 500-ns unrestrained simulation of the complex showed little movement between the Lec and YD shell (root-mean-square deviation [RMSD], ∼0.2 nm), suggestive of a stable complex (Figure S1G). In contrast, the NHL domain drifted from its original position at the YD shell (RMSD, ∼1.2 nm), suggesting that it is connected via a flexible hinge. The FN plug domain, ABD, and Tox-GHH domains were displaced by ∼0.3 nm RMSD. These results are consistent with previous cryoelectron microscopy (cryo-EM) studies, which also suggested flexibility in these areas. We also performed a 50-ns simulation in which the protein backbone atom positions were restrained to refine the positions of the side chains. The results suggested two binding hotspots within the interface; one of these contacts is centered around D55, located in the loop between Lphn2 Lec strands β2 and β3 (Figures 1F–1H, and S1H). Analysis of the hydrogen bond pattern showed that D55 forms a stable salt bridge with R2197, located in the top tier of the Ten2 YD shell. A second contact is formed by the β1-β2 and β4-h1 loops of Lec, which are binding to the lower part of the Ten2 YD shell. Here, too, negatively charged residues (Lphn2 D68 and D73) bind a positive patch on Ten2 (R1893 and K1866) (Figures 1F, 1G, 1I, andS1I). Analysis of the hydrogen bonding patterns within the 500-ns unrestrained simulation revealed a similar pattern of hydrogen bonds (Figures S1J–S1L).

5We produced a panel of Teneurin, Latrophilin, and FLRT constructs (Figures 2A–2C) and performed surface plasmon resonance (SPR) and cell-based binding experiments to validate our structural and simulation results. All relevant proteins were produced in mammalian HEK293 cells. We purified soluble ectodomains using affinity and size exclusion chromatography. TM constructs were tested for successful cell surface expression by immunostaining (Figure S2A;,, ). We also created “non-binding” mutants using an established method of introducing N-linked glycosylation sites at amino acid positions that are centrally located in the binding site. We refer to these mutants as “non-Latrophilin-binding Teneurin” (TenLT) and “non-Teneurin-binding Latrophilin” (LphnTL), in analogy to our previously published FLRTLF mutants, the “non-Latrophilin-binding FLRT” proteins. In addition, we produced a multiple-point mutant of Lphn1 that is based on our MD simulation results and does not depend on N-linked glycosylation: Lphn1TL2 (L39A, P51G, D54A, D67A, and D72A). A cell-based binding assay, in which TM Teneurin or Latrophilin constructs are expressed on HEK293 cells and incubated with soluble Latrophilin or Teneurin ectodomains, showed that Ten2LT, Lphn1TL, and Lphn1TL2 have indeed lost their ability to interact (Figures 2D–2F), validating the crystal structures. We also tested the mutants using SPR with consistent results (Figure 2G). The non-FLRT binding LphnFL mutant still binds Teneurin (Figure S2B). To test the mutations in trans-cellular adhesion assays, we subjected them to a previously described cell aggregation assay where Teneurin-expressing cells adhere to Latrophilin-expressing cells. We show that Ten2LT and Lphn1TL do not promote K-562 cell adhesion (Figures 2H and 2I). The specificity of the mutants is given by the single point mutations that were used to generate them, and so they are fully functional otherwise. For example, Lphn1TL cannot bind Teneurin, but it maintains its FLRT-binding capabilities. Therefore, Lphn1TL-expressing cells aggregate with FLRT2-expressing cells but not with Ten2-expressing cells (Figures 2H and 2I).

6Our structural results suggest that a ternary complex of Teneurin-Latrophilin-FLRT could form via the described 1:1 binding surfaces of Lec-Ten2 and Olf-FLRT (Figure 1C). We tested the formation of such a ternary complex using SPR. Pre-mixed Latrophilin and Teneurin proteins produced an increased binding response to FLRT protein compared with Latrophilin alone. Ten2 binds all three Latrophilins (Lphn1–Lphn3; Figure S2C) but, on its own, does not bind FLRT (Figures 2J, S2D, and S2E). Therefore, this increase must be due to Teneurin interacting with FLRT via Latrophilin. A cell-based assay confirmed these binding results. We showed that Teneurin only binds to cell surface FLRT when wild-type Latrophilin is present and not when a non-binding Lphn mutant is used (Figure 2K). We also used a pull-down assay where cells expressing full-length FLRT and Teneurin were mixed with cells expressing full-length Latrophilin or with non-transfected control cells. Immunoprecipitation of Teneurin lead to efficient pull-down of FLRT when the cells were mixed with Latrophilin-expressing cells, but not when they were mixed with control cells (Figures S2F and S2G). These data confirm that the Latrophilin extracellular domain can bind both Teneurin and FLRT simultaneously in vitro. The findings are consistent with the coincident binding model of Latrophilin, Teneurin, and FLRT, proposed by others in synaptic development. A summary of the specific mutations we used here is given in Figure 2L.

7We previously showed that Latrophilin and FLRT proteins affect embryonic cortical neuron migration in vitro and in vivo. Teneurins are expressed widely across the brain during development, including in the cortex, where they could play a role in pathfinding. Here we asked whether Latrophilin-Teneurin binding regulates embryonic cortical migration. In situ hybridization (ISH) for Latrophilins revealed that Lphn1 and 2 are expressed in neuron-enriched layers (cortical plate [CP] and intermediate zone [IZ]) and the apical progenitor (AP)-enriched layer (ventricular zone [VZ]), where radial glial (RG) cell bodies are located, from embryonic days 13.5 [E13.5] to E17.5 (Figures 3A, 3B, S3A, S3B, S3D, and S3E). This finding is consistent with single-cell RNA profiling data from E14.5 mouse cortex that also revealed expression of Lphn1 and Lphn2 in neurons and APs (Figure 3C). Combination of ISH with staining for the phosphorylated form of vimentin (Pvim), which labels dividing RGs, and the neuronal marker Ctip2 showed that Lphn1 and Lphn2 are expressed in neurons and RGs (Figure 3D). ISH for all four mouse Teneurins showed that these are predominantly expressed in neuron-enriched layers (CP and IZ) (Figure 3E; Figures S3C and S3F). In agreement with these results, analysis using data from two RNA-seq databases showed that Ten2 and Ten4 are highly expressed in neurons compared with APs (Figure 3F; ). Some Ten4 expression is also detected in RGs (Figure S3G; ). These results were further confirmed by co-staining with the neuronal marker Ctip2 and the RG marker Pvim. Ten2 and Ten3 showed reduced staining in APs compared with neuron-enriched layers (Figure 3G).

8Given that FLRTs are also enriched in neurons during cortical development (;; Figure S3H), we asked whether FLRTs and Teneurins are co-expressed in migrating cortical neurons. Single-cell RNA profiling analysis showed a positive correlation between Ten2 and FLRT1/3 and Ten4 with FLRT2 in cortical neurons (Figure S3I). Co-expression of Ten2 and FLRT3 was strongest in migrating neurons compared with other cell types (Figure S3J). Surface staining for both Ten2 and FLRT3 revealed that both receptors are located in close proximity on the cell body and neurites/growth cones of cortical neurons. Addition of Lphn1 (Lec-Olf) protein induced more proximal localization of FLRT3 and Ten2 compared with addition of the Lphn1 (Lec-Olf) non-Teneurin and non-FLRT-binding (TL-FL) mutant protein. The mutant protein also bound less to cultured neurons compared with the wild type, and it co-localized less with Ten2 and FLRT3 staining (Figures 3H, 3I, and S3K–S3M). Latrophilin expression in cis can moderate but does not abolish Teneurin or FLRT binding to externally presented Latrophilin (Figures S3N–S3P).

9In agreement with these results, our crystallography, SPR, and cell-binding experiments suggest that Latrophilin, FLRT, and Teneurin physically interact (Figures 1 and 2). Pull-down experiments using E15.5 mouse brain lysate show that Ten2 and Latrophilins co-immunoprecipitate with FLRT, suggesting that the three proteins also interact in vivo (Figures 3J, 3K, S3Q, and S3R). Based on these results, we developed a working model in which migrating neurons co-expressing Lphns, Teneurins, and FLRTs interact with Lphns present in RG cells and/or other cortical neurons (Figure 3L). We proceeded with functional analysis to understand the roles of these proteins in early cortical development.

10To study the effects of Latrophilin-Teneurin binding on cortical neuron migration, we performed time-lapse imaging of E15.5 embryonic cortical explants grown on Lphn1 (Lec-Olf)-coated dishes and measured the migration of neurons exiting the explant (Figures 4A and S4A). Using automatic tracking (Video S1), we found that cortical neurons migrated slower and shorter distances on Lphn1 (Lec-Olf)-coated surfaces compared with control surfaces (Figures 4B and S4B). Similar effects were observed with the TL-FL Lec-Olf single mutants (Figures S4C and S4D; Video S1). In contrast, neurons migrating on the double-mutant protein (LphnTL-FL Lec-Olf) behaved similarly as those on control protein.

11Given that Lphn1 is highly expressed in RG cells, whereas its binding partners Teneurins and FLRTs are present in migrating neurons (Figures 3 and S3), we addressed the complexity of Lphn1 function in the context of neuron-RG cell interactions. Cortical migration relies on a delicate balance between neuron-neuron and neuron-RG fiber interactions. Indeed, altering glial-guided neuronal migration affects neuronal dynamics and morphology as well as their tangential dispersion.

12We chose arrays of parallel aligned nanofibers to mimic the fibrillary environment of RG cells. As a source of neurons, we used cortical explants that, after being positioned onto nanofibers, displayed directed axon growth and cell migration along the fibers (Figures 4C and S4E). Neurons exiting the explants displayed similar features as observed in vivo, such as transitioning from multipolar to bipolar morphology prior to migrating along nanofibers and transient branching of their leading process, as described recently (; Figure 4D; Video S2). Neurons occasionally switched between neighboring fibers (Figure 4E) but produced mostly linear paths, as observed in vivo (Figure 4F; Video S2). Using semi-automatic analysis (Figure S4E), we found that neurons migrated shorter distances on nanofibers coated with Lphn1 (Lec-Olf) compared with FC control protein (Figures 4G, 4H, and S4F). Non-Teneurin (TL) and non-FLRT-binding (FL) Lphn1 (Lec-Olf) mutant proteins were equally effective in slowing neuron migration. The double mutant Lphn1TL-FL (Lec-Olf) resulted in reduced effects on neuron migration compared with the wild-type protein, indicating that additive or coincident binding of Teneurins and FLRTs affects neuron migration. The fact that this mutant produces partial rescue of the migration effect suggests that other interactions may also play a role; e.g., a weak glycan interaction site has been reported on mLphn1 Lec, with Kd values in the millimolar range. The mean axon length and the percentage of explants producing migration were similar under all conditions (Figures 4I, 4J, S4F, and S4G). In summary, these results indicate that Lphn1 delays cortical cell migration by binding Teneurins and FLRTs while having no effect on their axon growth.

13The mechanism by which cell surface proteins decrease cell migration speed could be due to a change in the balance of adhesion versus repulsion. Modulating either can lead to a reduction of migration in other systems. For example, during Xenopus gastrulation, migrating cells alternate between attachment and detachment. Reducing contact repulsion, in that case mediated by ephrinB-EphB signaling, increases attachment and reduces cell motility. Conversely, increasing ephrinB-EphB repulsion induces cell detachment, also affecting migration. Using stripe assays, we had shown previously that Lphn3 (Lec-Olf) repelled embryonic cortical neurons. Here we found that Lphn1 (Lec-Olf) is also mildly repulsive for cortical neurons (Figures 5A, 5B, and S5A). Similar to the nanofiber assay described above, using stripe assays, we find that non-Teneurin (TL) and non-FLRT-binding (FL) Lphn1 single mutants are also repulsive. The Lphn1TL-FL double mutant produces no response compared with the control. Time-lapse imaging of dissociated cortical neurons and GFP electroporated neurons choosing between alternate stripes of Lphn1 (Lec-Olf) and the TL-FL mutant showed that cell bodies and small neurites prefer the mutant (Figures 5C and 5D; Videos S3 and S4). These data suggest that the Lphn1-induced repulsion from stripes is due to its interaction with Teneurins and FLRTs in trans. Lphn1 was not repulsive for cortical axons (Figures 5E, 5F, and S5B), which is also in agreement with the nanofiber assay. Also, GFP electroporated axons were not repelled by Lphn1 in time-lapse experiments. The growth cones of these axons did not show a preference for the Lphn1 (Lec-Olf) wild type or the TL-FL mutant (Figures 5G and 5H; Video S4). In summary, these results suggest that Lphn1 is repulsive for migrating cortical cells through an interaction with Teneurins and FLRTs but not for their axons.

14Having established that Lphn1 binding of Teneurins and FLRTs is repulsive for migrating neurons in vitro, we next addressed their function in the developing cortex. Using in utero electroporation (IUE) at E15.5, we overexpressed full-length Ten2 plus GFP, the non-Latrophilin-binding Ten2LT plus GFP, or GFP control in cortical cells and analyzed their distribution at E18.5 (Figure 6A). We confirmed overexpression of Ten2 and Ten2LT in migrating neurons (Figure 6B) and their correct localization and ligand-binding abilities at the cell surface (Figures S6A–S6C). We found that Ten2 overexpression in cortical migrating neurons delayed their migration. This delay was not observed when overexpressing Ten2LT (Figures 6C and 6D), suggesting a functional role of Teneurin-Latrophilin interaction. Similar results were obtained when targeting cortical neurons at earlier stages (E13.5) (Figures S6D and S6E), which is consistent with the early expression of these proteins (Figures S3A–S3C). We performed live imaging of embryonic cortex ex vivo to find out whether Ten2 overexpression affects cortical migration. For these experiments, brains were sliced and imaged 48 h after electroporation with Ten2/GFP or Ten2LT/GFP and mCherry to also label control neurons. Tracking of migrating neurons revealed that Ten2-expressing neurons migrated slower and remained longer in the IZ compared with control neurons (Figures 6E–6G; Video S5). The effects on neuronal migration were more modest when overexpressing Ten2LT mutant protein (Figures 6E–6G; Video S5). In a separate set of experiments, we knocked down endogenous Ten2 in E15.5 cortices by embedding small hairpin RNA (shRNA) target sequences into the pCAG-miR30 vector system (; Figures 7A–7C, S7A, and S7B). Analysis at E18.5 showed reduced migration of these neurons (Figures 7D and 7E). Similar results were obtained when Ten2 expression was targeted by CRISPR using the pX458 system (; Figures S7C and S7D). We also expressed a secreted version of Lphn1 (Lec-Olf) to compete with its endogenous ligands at E15.5, and this produced a strong delay in neuronal migration. This effect was not observed when using the mutant (TL-FL) construct (Figures 7F, 7G, S7E, and S7F). These results show that Latrophilin interactions are essential for cortical neuron migration in vivo (Figure 7H).

15Teneurins are highly conserved, with chicken and murine Teneurin 2 sharing 93% and 99% sequence identity with human Teneurin 2, respectively. The structural mechanism of how Teneurin binds to Latrophilins has long remained enigmatic, and previously available knowledge was lacking detailed structural information. The data presented here demonstrate that the Latrophilin-binding site on Teneurin comprises separate adjacent binding motifs located across the tiers of the spiraling β-sheet of the YD shell. Superposition with the FLRT2 LRR, as found previously bound to Lphn3, gives a compact ternary model that still involves the Teneurin YD shell domain rather than other Teneurin domains. The interface between the two bottom tiers of the Teneurin YD shell is of structural interest because it enables passage of the internal linker through the YD shell and into the ABD. In the related bacterial TcB/TcC toxins, the equivalent interface area between the tiers of the YD shell is formed by two separate subunits of the toxin protein, suggesting that they come together post-translationally in bacteria. Given its geometry, this area of Teneurin must undergo significant conformational changes during protein folding and, therefore, may be inherently mobile. Latrophilin binds across these tiers of the Teneurin YD shell like a clamp and appears to stabilize the fold of this structural module.

16Previous studies have emphasized that the inclusion of an alternatively spliced loop in the Teneurin NHL loop regulates Teneurin interactions, with some reports suggesting that inclusion of the loop inhibits Latrophilin binding. Others found that the presence of this loop does not affect Teneurin-Latrophilin binding but that it promotes homophilic interactions of Teneurin together with an alternatively spliced sequence in the Teneurin EGF domains. Our Teneurin constructs include the alternatively spliced loop in the NHL domain but not that in the EGF domains. In agreement with, the constructs do not promote homophilic interaction, and they bind Latrophilins. The crystallographic data explain these binding results; the NHL domain does not engage with Latrophilin directly. Using the new structural data, we revealed a molecular map of how Teneurin, Latrophilin, and FLRT termini interact in vitro and designed a toolkit of mutants that control the individual interactions formed by these receptors. These mutant tools are used to interrogate the functions of specific interactions in cellular and in vivo settings.

17Teneurins have recently emerged as key regulators of synaptic wiring. This highly evolved function contrasts with the evolutionarily early origin of Teneurins in unicellular organisms that lack a nervous system. In these organisms, Teneurin-like proteins likely engage in other types of cell-cell or cell-extracellular matrix interactions. This is consistent with studies in worms, where Teneurin (Ten-1) is essential for gonad epithelialization and basement membrane integrity.

18Here we show that, in the mammalian nervous system, Teneurins play important roles at earlier time points, much before synapse development. They direct the migration of embryonic cortical neurons, possibly by regulating their interaction with the radial glia scaffold. Although Teneurins and FLRTs are mainly expressed in migrating neurons, Latrophilins are expressed in neurons and RG cells. Mechanistically, the migrating neuron co-expresses Teneurins and FLRTs, and these two proteins could therefore bind Latrophilins coincidently in trans on opposing RG cells or other neurons (Figure 7H). This configuration is similar to the one proposed for their synaptogenic function: Teneurins and FLRTs on the pre-synaptic side interact with Latrophilins in trans on the post-synaptic side. Other configurations are conceivable, and it is possible that different context-dependent complexes form transiently as the cells migrate through their complex environment. Cell type-specific manipulation of expression of Teneurins, FLRTs, and Latrophilins will be needed to unravel the most important configurations for cortical neuron migration in vivo. Given that the formation of synapses is an adhesive/attractive process, we were surprised to find that the interaction between Teneurins, FLRTs, and Latrophilins during cell migration is repulsive. The delay in cell migration induced by overexpression of Teneurin could be caused by loss of traction rather than an increase in adhesion. Such a dual role in repulsive cell guidance and synaptogenesis is not without precedent. Erythropoietin-producing human hepatocellular (Eph) receptor/ephrin signaling has long been reported to mediate these processes, but despite intense research, the underlying signaling mechanisms that convert repulsion into synapse formation are poorly understood. Previous studies have identified different molecules promoting adhesion of migrating neurons to RG fibers such as connexin26/43, focal adhesion kinase (FAK), and N-cadherin. However, the molecules that mediate repulsion between neurons and RG fibers remain largely unknown.

19Given that the synaptogenic functions of Teneurins and FLRTs require their localization on the pre-synaptic side, we were surprised to see that cortical axons showed no response toward Latrophilins. The easiest explanation would have been that, during the migratory phase, Teneurins and FLRTs are excluded from the axonal compartment and that these proteins are induced in axon terminals during synaptogenesis. However, we find that Ten2 and FLRT3 are co-expressed and show uniform distribution on the cell surface during the migration phase. These results suggest that the differential responses—i.e., repulsion of somata and small neurites and adhesion of axon terminals—are the result of differences in the downstream signaling pathways between the somatodendritic and axonal compartments. Similar results have been shown for Semaphorin 3A, which is attractive for cortical dendrites but repulsive for axons because of asymmetrically localized guanylate cyclase. The highly polarized structure of migrating neurons could also contribute to the different response between its dendrites and axons. The leading process of migrating neurons seems to preferentially mediate the interactions toward RG fibers. These increased contacts induce polarization of downstream signaling molecules, such as RhoA, being recruited to the leading process and Rac1 to the opposed neurite that will become the axon.

20In vivo, overexpression and knockdown of Ten2 in cortical neurons delayed their migration toward the CP, and this effect was not observed when overexpressing the Ten2 mutant defective in Lphn binding. The impairing effect of tampering with the Ten2 levels on migrating neurons is reminiscent of other receptors regulating cell migration, such as Neuroligin 2. FLRT loss-of-function and gain-of-function experiments also resulted in abnormal migration of cortical neurons. The need for finely balanced levels of Ten2 in migrating neurons is a likely reason why attempts at loss-of-function rescue experiments, in which we tried to simultaneously knock down endogenous Ten2 and overexpress a resistant version by IUE, failed. It is likely that the levels of resistant Ten2 overexpressed in these neurons is high, more similar to our overexpression experiments than the correct endogenous levels. We complement our loss-of-function results with a separate in vivo assay where we overexpressed the secreted portion of the Latrophilin ectodomain that interacts with FLRTs and Teneurins (Lec-Olf) to effectively compete with the interactions of endogenous Lphn with its binding partners and delay cell migration. Regarding the downstream signaling pathways that mediate repulsion, very little is known. We have shown previously that Lphn-FLRT and FLRT-Unc5 interactions can trigger repulsive responses in cortical neurons in vitro. In future work, it would be interesting to assess the possible role of Unc5 receptors in the context of a Teneurin-FLRT-Latrophilin complex. Unc5 are classic repulsive cell guidance receptors and form supercomplexes with Latrophilins and FLRTs. Recruitment of Unc5 to a Teneurin-FLRT-Latrophilin complex may enhance repulsive signaling.

21In conclusion, our crystal structures revealed the main binding site for Latrophilin on the Teneurin YD shell domain. Latrophilin-Teneurin and Latrophilin-FLRT interactions give rise to a ternary complex and can be controlled by specific mutations in the Lec and Olf domains, respectively. Latrophilin binding of Teneurins and FLRTs, previously shown to engage neurons in synapse formation, slows down the migration of embryonic cortical neurons by contact repulsion. In our paradigm, the repulsive response involves the somatodendritic compartment of neurons without affecting their axon growth and extension, which are initiated concomitantly with migration. Teneurins and Latrophilins show broad expression during brain development, and so the reported mechanism could also be involved in controlling neuronal migration in other brain regions.

22Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Elena Seiradake (elena.seiradake@bioch.ox.ac.uk). We are glad to share plasmids published in this study with reasonable compensation by requestor for its processing and shipping.

23All mice (C75BL/6 background) were housed with 12:12h light/dark cycle and food/water available ad libitum. All experimental procedures involving mice were conducted in accordance with regulations set by the government of Upper Bavaria (License number 55.2-1-54-2532-57-2015). All histological experiments were carried out between E15.5 and E18.5.

24Cortical explants from E15.5 embryos were dissected out and placed on 60-mm dishes or nanofibers. Coating was performed using 50 μg/ml of FC or Lphn1 proteins and 20 μg/ml of laminin (surface experiments) or 40 μg/ml of FC or Lphn1 proteins with 100 μg/ml poly-D-lysine and 20 μg/ml of laminin (nanofibers). Neurons were dissociated from cortices of E15.5 embryos and cultured on stripes (same coating as explants) or coverslips coated with 0.5mg/ml Poly-D-Lysine in 24 well-plates. Explants and neurons were cultured for 1-3 days in vitro at 37°C, 5% CO2 in Neurobasal medium supplemented with B27. Samples were fixed with 4% Paraformaldehyde for 10 min and processed for immunostaining.

25HEK293T cells (ATCC, cat# CRL-3216) and HEK293S cells (ATCC, cat#CRL-3022) used in this study were cultured in Dulbecco’s Modified Eagle Medium, DMEM, supplemented with 10% FBS, 1% NEAA and 1% L-Glutamine at 37°C and 5% CO2. K-562 cells (ATCC, cat# CCL-243) used in this study were cultured in phenol red-free Iscove’s Modified Dulbecco’s Medium, IMDM, supplemented with 10% FBS, 1% NEAA and 1% L-Glutamine at 37°C and 5% CO2. Protein expression was induced by transfection with plasmid DNA and polyethylenimine as the transfection reagent.

26The constructs used in this study (Figures 2A–2C) were cloned into pHLSec or pCAGIG (Addgene, 11159) variant vectors with relevant tags for expression in cell lines or neurons, respectively. For protein purification, we used pHLSec vectors which also code for a C-terminal 6xHis-tag, for SPR we used a C-terminal Avi-tag, and for visualization on HEK or K-562 cells we used pHLSec variants that add an intracellular mVenus, mRuby, and/or an extracellular Flag, Myc and/or HA tag as indicated. For cloning of Ten2LT (1892N+K1894T) we used forward and reverse primers, ctacgatgataaccgcacattcaccctgaggata and cagctgatcgtaaattatcctcagggtgaatgtgcggttatcatcgtag, respectively. For cloning of Lphn1TL (E39N) we used forward and reverse primers, ctatcctgtgagggttattctatagacc and accctcacaggatagatttcgtctaac, respectively. For cloning of Lphn1TL2 (L39A, P51G, D54A, D67A, D72A) we used forward and reverse primers, gccgaccctttccagatggagaatgtgcagtgctacttgcctgacgc and ctggaaagggtcggcagcgcagatcttgtcagctgtgcg, respectively. We also used constructs that were from previous papers, Lphn1FL and FLRTLF. In detail, the forward and reverse primers that were used for cloning Lphn1FL (R292N+R294T) were cctgttcttcaacaaagagaataccacaaacatcgtgaagttcgacctgc and gcaggtcgaacttcacgatgtttgtggtattctctttgttgaagaacagg, respectively. The forward and reverse primers used for cloning FLRTLF (R186N+D188T) were gggcttcctgtagacttgcaagagctgaatgtgactgaaaaccgaattgccgtcatatc and gatatgacggcaattcggttttcagtcacattcagctcttgcaagtctacaggaagccc, respectively.

27Unless indicated otherwise, all purified proteins were expressed in adherent HEK293 cells following established procedures. More specifically, for each protein preparation, three liters of HEK cell culture were grown in Dulbecco’s Modified Eagle Medium, DMEM (Life Technologies, cat#21969035), supplemented with 10% FBS (Life Technologies, cat# 10270-106), 1% NEAA (Life Technologies, cat# 11140035) and 1% L-Glutamine (Life Technologies, cat#25030-024) at 37°C and 5% CO2. K-562 cells (ATCC, cat# CCL-243). Prior to transfection, the FBS content was reduced to 2%. The cells were transfected with 6 mg of purified plasmid DNA, previously mixed in a 1:2 (mass) ratio of DNA: polyethyleneimine (cat# 208727, Sigma Aldrich) as transfection reagent. The medium was harvested 5-10 days later, buffer-exchanged to 1x PBS (Sigma) supplemented with 150 mM NaCl and 20 mM Tris, pH 7.5. The proteins were purified by loading onto a 5ml HisTrap HP column (GE Healthcare). The column was then washed with 30 column volumes of buffer containing 40 mM imidazole, and the protein eluted using buffer containing 500 mM imidazole. The proteins were then purified using a Superdex200 size exclusion column (GE Healthcare) using buffers containing 150 mM NaCl and 20 mM Tris, pH 7.5. The proteins used for crystallization were expressed under conditions that reduce glycan heterogeneity. Ten2 and Lphn1 (Lec-Olf) proteins were expressed in HEK293T cells in the presence of kifunensine, and Lphn2 Lec was expressed in HEK293S GlnTI- cells. Prior to crystallization, the relevant proteins were mixed together in a 1:1 molar ratio, both complexes were concentrated to 4.9 mg/ml and mixed in 1:1 volume ratio with crystallization solution. The Ten2 and Lphn2 Lec complex was mixed with 0.1 M calcium acetate, 0.1 M sodium acetate, pH 4.5, 10% (w/v) PEG 4000, and crystals grew in sitting drops at 4°C before flash-freezing in liquid nitrogen in reservoir solution supplemented with 25% glycerol. The Ten2 and Lphn1 (Lec-Olf) complex was mixed with 0.1 M potassium chloride, 0.1M HEPES pH7.5, 15% (w/v) PEG6000, and crystals grew in sitting drops at 18°C before flash freezing in reservoir solution supplemented with 25% glycerol.

28We performed the X-ray diffraction data collection at the Diamond Light Source beamline I03, in each case from a single crystal at 100 K. Data from the Ten2-Lphn2 complex crystals were integrated data up to 3.6 Å resolution using xia2 and we solved the structure by molecular replacement using phaser with mouse Lphn3 Lec (PDB ID 5AFB; ) and chicken Ten2 (PDB ID 6FB3; ) as inputs. We placed two copies of each Teneurin and Latrophilin in the asymmetric unit of the P1 cell. After placing Lphn3 Lec, we substituted this with a mouse Lphn2 Lec domain homology model, generated by SWISS-MODEL. Glycans and selected solvent atoms that were present in the high-resolution molecular replacement models were included as appropriate. Given the higher resolution data used to refine the input models of our molecular replacement job, we targeted the model to these structures during refinement. We also used automatic non-crystallographic symmetry restraints, and TLS refinement. Data for the Ten2-Lphn1 complex was integrated to 4 Å resolution using xia2 and we solved the structure by molecular replacement using phaser with the Ten2-Lphn2 structure presented here, and a homology model of the Lphn1 Olf domain, generated by SWISS-MODEL. Again, two copies of the complex were found in the asymmetric unit. We refined both structures using programmes in ccp4, buster and coot. Validation of the models was performed in molprobity. We created figures using pymol. Sequence conservation analysis was performed with Consurf. For this analysis, sequences of all homologs were aligned from Mus musculus, Xenopus tropicalis, Gallus gallus and Danio rerio using Clustal Omega. Crystallographic details are summarized in Table S1.

29We followed the same protocol as previously described to refine the interaction surfaces of X-ray crystallography-derived complexes. In detail, Molecular dynamics simulations were performed using GROMACS 2018 with an AMBER99SB force field. We embedded the complex model in a box of water (SPC/E water model; ). Na and Cl ions were added up to 150 mM. Energy minimization was performed using the steepest descent algorithm and each system was equilibrated with a constant temperature (canonical ensemble, NVT, 310 K) ensemble for 100 picoseconds (ps), followed by a 100 ps equilibration at constant pressure (isothermal- isobaric, NPT, 1 bar). For equilibration and production runs, we applied the velocity-rescaling thermostat on protein and solvent, coupled with the Parrinello–Rahman barostat, with a time constant of 2.0 ps and compressibility of 4.5x10−5 bar-1. Long-range electrostatics were modeled using the Particle-Mesh Ewald method. All bonds were treated using the LINCS algorithm. The integration time step was 2 femtoseconds. We performed two different simulations: (i) 500 ns of unrestrained simulation to assess the stability of the Ten2-Lphn2 Lec complex; and (ii) 50 ns semi-restrained simulation. For the latter, the Teneurin backbone was constrained and we applied distance restraints on the Latrophilin backbone. By doing this we allowed Latrophilin to move with respect to Teneurin while keeping the backbone of both proteins in the configuration determined by X-ray crystallography. Side chains were allowed to move freely in this simulation. The restrained simulation was used to analyze H-bonds formed at the protein interfaces throughout the last 40 ns of simulation using the MDAnanlysis library (relevant Jupyter Notebook and tcl script available at: https://github.com/MChavent/Hbond-analysis). We used a donor-acceptor distance cut-off of 3.0 Å and a cut-off angle of 120°. The stability was defined as correlating with the percentage of the simulated time in which an atom forms the H-bond with its partner

30Equilibrium binding experiments were performed at 25°C using a Biacore T200 instrument (GE Healthcare) using PBS + 0.005% (v/v) polysorbate 20 (pH 7.5) as running buffer. The regeneration buffer was 2 M MgCl2. Latrophilin constructs were biotinylated enzymatically at a C-terminal Avi-Tag and coupled to a streptavidin-coated CM5 chip (GE healthcare, cat# 29104988). Data were analyzed using the BIAevaluation software (Biacore, GE Healthcare). Indicative KD and Rmax values were obtained by nonlinear curve fitting of a 1:1 Langmuir interaction model (bound = Rmax/(KD +C), where C is analyte concentration calculated as monomer (Table S2).

31HEK293 cells were transiently transfected with transmembrane constructs of Teneurin or Latrophilin that were fused to an intracellular mVenus-tag and an extracellular HA, Flag or Myc tag. Depending on the tag included, anti-HA (SIGMA-Aldrich, Cat No. H3663), ant-Flag (SIGMA-Aldrich, Cat No. F1804) or anti-Myc (Abcam, Cat No. ab19233) were added to the cells at a concentration of 5 μg/ml. Cells were incubated for 1 hour on ice and washed with phosphate buffer saline (PBS, Lonza). Cells were fixed in 4% PFA (20min, on ice), immunostained with Cy3-labeled anti-mouse antibody (Abcam, Cat No. ab97035) or anti-chicken A568 antibody (Abcam, Cat No. ab175711) and mounted. Non-transfected cells acted as negative control.

32HEK293 cells were transiently transfected with the same constructs described above. His-tagged purified proteins were pre-clustered with 6x-His Tag antibody (Life Technologies, Cat No. 372900) for 1hr on ice and were added to the cells at a concentration of 5 or 10 μg/ml, for purified Lphn or Ten2 ectodomain samples, respectively. Cells were washed with phosphate buffer saline (PBS) and fixed in 4% PFA (20min, on ice), immunostained with Cy3-labeled anti-mouse antibody (Abcam, Cat No. ab97035) or anti-chicken A568 antibody (Abcam, Cat No. ab175711) and mounted. Images were analyzed using ImageJ. For each image, the integrated density (the sum of the values of the pixels in the image) corresponding to bound protein was quantified (red) and divided by the integrated density corresponding to mVenus overexpression (green).

33In case of experiments where HEK293 expressing FLRT3 were incubated with purified Lphn1 (Lec-Olf) or Ten2 ectodomain. HEK293 cells were transfected with a pcDNA3 vector (Invitrogen) containing full-length mouse FLRT3 with a C-terminal FLAG-tag. 100 nM Ten2 or Lphn1 (Lec-Olf) were pre-clustered with 50 nM Anti-6XHis-tag FITC (Abcam, ab3554) or Anti-6XHis-tag DyLight 650 (Abcam, ab117504) antibodies for 1hr at room temperature and were added to the cells simultaneously. Following an incubation for 20 min at room temperature, cells were washed with PBS. Cells were then fixed in 4% PFA for 10 min at room temperature, permeabilized in PBS containing 0.1% Triton (CarlRoth) and 1% bovine serum albumin (Sigma) and immunostained for FLRT3 overexpression with an anti-FLAG antibody (1:1000, Sigma, F9291).

34K-562 cells were cultured according to the manufacturer’s instructions using Iscove’s Modified Dulbecco’s Medium without phenol red supplemented with 10% FBS. Cells were transfected with pHLSec plasmids coding for Teneurin (N-terminal mVenus) or Lphn (C-terminal mRuby) as indicated, using Lipofectamine LTX with PLUS reagent (Life Technologies, 15338100). 24 hours after transfection, the two population of cells were mixed in 24-well plates and incubated at 37°C, 5% CO2 and 250 rpm for ∼30 min.

3550 μg/ml His-tagged protein was mixed with 120 μg/ml Cy3-conjugated αFc (Life Technologies A11014) in PBS. Matrices (90 μm width) were placed on 60 mm dishes and proteins injected. After 30 min incubation at 37°C, dishes were washed with PBS and matrices removed. Dishes were coated with 50 μg/ml Fc protein mixed with 120 μg/ml anti-hFc (Jackson ImmunoResearch 109-005-098) for 30 min at 37°C and washed with PBS. Stripes were further coated with 20 μg/ml Laminin in PBS for at least 2 hours and washed with PBS. Cortical neurons (E15.5) or explants (E15.5) were cultured on the stripes in Neurobasal medium supplemented with B27 (Invitrogen) and, in case of explants, 0.4% methyl-cellulose (Sigma). After 24 (neurons) or 48 (explants) hours and fixed with 4% PFA in PBS for 20 min at room temperature (RT). Neurons and explants were washed and incubated with rabbit monoclonal anti-beta-III tubulin antibody (Sigma) after 20 min permeabilization in 1% BSA, 0.1% Triton X-100/PBS. Cy2 anti-rabbit IgG secondary antibody (Jackson ImmunoResearch, cat#111-225-144) was used to visualize the tubulin signal. Nuclei were counterstained with DAPI before mounting. The numbers of beta-III-tubulin+ or DAPI+ pixels on red or black stripes were quantified using ImageJ (version 1.51p).

36In utero electroporation (IUE) was performed at E13.5 or E15.5 on anesthetized C57BL/6 mice, as decribed below. DNA plasmids were used at 2 μg/μl and mixed with 1% fast green (Sigma, final concentration 0.2%). Knockdown of Ten2 by shRNA used the following sequences embedded in pCAG-miR30 plasmid: shRNA1: CTCCCTGTACGTTCTGGAGAAC, shRNA2: TTCACATACAGCGCTGACAATG. CRISPR sequences embedded in pX458 plasmids were: CRISPR#1: AGCTCCTGCGCCCAACGACC CRISPR#2: AACAGTAACCACTTCGGTGT. Scrambled sequence as control was: GCTCGCACACTAGCCGCCAC. Plasmids were injected into the ventricle with a pump controlled micropipette. After injection, six 50 ms electric pulses were generated with electrodes confronting the uterus above the ventricle. The abdominal wall and skin were sewn and the mice were kept until the desired embryonic stage. The pCAG-miR30 plasmids were validated in HEK293T cells, by co-transfection with mouse Ten2 (ecto) construct or chicken Ten2 (ecto), which acted as a non-sensitive control. The expression of Ten2 was analyzed by western blotting, using mouse anti-His antibody (QIAGEN, 34660). Secreted versions of Lphn1(Lec-Olf) and Lphn1 (Lec-Olf) TL-FL used in IUE experiments were cloned into the pCAGIG vector, and encoded with a C-terminal Myc tag. The expression of these constructs were validated by expression in HEK293T cells and analyzed on western blots using chicken anti-Myc antibody (Abcam, ab19233).

37Embryos were electroporated at E15.5 as described above using pCAG-mCherry (control neurons) together with pCAG-Ten2-IRES-GFP or pCAG-Ten2LT-IRES-GFP. After 48 hs, embryonic brains were dissected in ice cold sterile filtered and aerated (95% O2/5% CO2) dissection medium (15.6 g/l DMEM/F12 (Sigma); 1.2 g/l NaHCO3; 2.9 g/l glucose (Sigma); 1% (v/v) penicillin streptomycin (GIBCO, cat#15140148)). Brains were embedded in 4% low melting agarose (Biozym, cat#840101) and cut into 300μm thick sections using a vibratome (Leica, VT1200S). Sections were suspended in a collagen mix (64% (v/v) cell matrix type I-A, Nitta Gelatin; 24% (v/v) 5 x DMEM/F12; 12% (v/v) reconstitution buffer (200mM HEPES; 50mM NaOH; 260mM NaHCO3) and transferred onto a cell culture insert (Millicell; PICMORG50). Sections were incubated for 10 min at 37°C to solidify collagen. 1.5 mL slice medium (88% (v/v) dissection medium; 5% (v/v) horse serum; 5% (v/v) fetal calf serum; 2% (v/v) B27 supplement (GIBCO, cat#17504044); 1% (v/v) N-2 supplement (GIBCO, cat#17502001)) was added into the dish surrounding the culture insert and incubated for 30 min at 37°C. Before start of time-lapse experiment, culture medium was added on top of the sections to allow objective immersion. Sections were imaged using a 20x water immersion objective on a Leica SP8 confocal microscope system equipped with a temperature-controlled carbon dioxide incubation chamber set to 37°C, 95% humidity and 5% CO2. Sequential images were acquired every 20 min for 14-60 hr. Single cell movement was tracked using the Fiji plugin “Manual Tracking.” Single cell track analysis and plotting was carried out using homemade python scripts.

38Fresh E15.5 mouse cortices were homogenized for 1min at 4°C with an electric homogenizer using the following lysis buffer: 50 mM Tris-HCL (pH 7.4), 150mM NaCl, 2mM EDTA, 1% Triton X-100 and protease inhibitors (Roche ref. 04693116001). Samples were incubated on ice for 20 min and centrifuged for 10 min at 3000 rpm. Supernatant was collected and protein was measured using the Bio-Rad protein assay (Biorad, 5000001). 1 mg of protein at a final concentration of 2 μg/μl in lysis buffer (volume: 500 μl) was used for each pull-down. Control pull-down contained lysate and 2 μg of goat anti-human IgG antibody (Jackson Immunoresearch, 109-005-098) while FLRT3 pull-down used 2 μg of goat anti-human FLRT3 antibody (R&D, AF2795). Samples were incubated overnight at 4°C under rotatory agitation. The next day, 10 μg of Sepharose beads were added to each sample (Amersham CL-4B, 17-01780-01, 50% v/v in lysis buffer) and incubated for 4 hours under rotatory agitation. Sepharose beads were centrifuged for 5 min at 3000 rpm and washed three times (first wash with 400 μl of lysis buffer, second wash with 1:1 (v/v) lysis buffer:PBS, last wash only PBS). Pulled-down samples were processed for mass spectrometry (MaxQuant run, Proteomic facility, Max Planck Institute of Biochemistry, Martinsried, Germany) or standard western blot. For western blotting, samples were run on 6% SDS gels, transferred to PVDF membranes and blotted for Ten2 (Novus Biologicals, rabbit, NBP2-55763) or FLRT3 (R&D, AF2795). For mVenus-tagged Ten2 pull-downs, adherent HEK293-T cells were transfected for 24 hours, with 0.5 μg pHLSec DNA coding for murine FLRT2 (Flag tagged) and 3 μg of pHLSec DNA coding for chicken Ten2 (mVenus tagged), per 2 mL of cell culture. Separately, cells were transfected with 3 μg of murine Lphn3, per 2 mL cell culture. 4 mL of each cell was used per experiment. Next day, cells transfected with Lphn3 were washed with PBS and incubated at 37°C with PBS supplemented with 1mM EDTA for 5min. The PBS/EDTA solution was then removed and cells were resuspended in 1ml DMEM media supplemented with 10% FBS, and pelleted at 100 x g for 4 min. Equal volumes of Lphn3-cells were added to adherent cells transfected with Ten2/FLRT2 and incubated at 37°C for 1hr. Cells were resuspended in ice-cold lysis buffer (1% Triton X-100, 50 mM Tris-HCl at pH7.5, 150mM NaCl and EDTA-free protease inhibitors (Sigma)). Cells were disrupted by mechanical force, and incubated on ice for 20 min. Cell debris was removed by centrifugation and expression control samples were taken at this point. The lysate was incubated with 2 μg/ml of rabbit anti-GFP antibody (Life Technologies, A11122) at 4°C for 1 hour under rotary agitation. 45 μl of a 50% suspension of pre-blocked Protein-G Sepharose 4 FF (Sigma, cat#P3296) was added per 1 mL of lysate, and incubated for 2 hours. After incubation, Sepharose beads were pelleted and washed twice with ice-cold lysis buffer, once with 1:1 ice-cold lysis buffer:PBS, once with ice-cold PBS, then resuspended in PBS and added with SDS-loading buffer. FLRT protein was detected using anti-Flag (Sigma, F1804) on a western blot. Band intensities were analyzed using ImageJ (Fiji) and normalized within each blot.

396 well-plate parallel nanofibers (700nm width, Sigma, Z759333-1EA) were coated with 40 μg/ml of specified protein (FC, Lphn1(Lec-Olf) and its mutant versions) and 100 μg/ml poly-D-lysine (Sigma) in PBS overnight (37°C, 65% humidity and 5% CO2). Next day, plates were washed with PBS and coated with 20 μg/ml laminin in PBS overnight. The next day, plates were washed with PBS and 36 cortical explants (E15.5) were placed per well (6x6 grid, see Figure S5E) and cultured for 2 days in Neurobasal medium supplemented with B27 (Invitrogen) supplemented with 0.4% methyl-cellulose (Sigma). Then the explants were fixed with 4% PFA for 20min, washed with PBS and incubated with rabbit monoclonal anti-beta-III tubulin antibody (Sigma) after 20 min of permeabilization in 1% BSA, 0.1% Triton X-100/PBS. Cy2 anti-rabbit IgG secondary antibody (Jackson ImmunoResearch, Cat#111-225-144) was used to visualize the tubulin signal. Nuclei were counterstained with DAPI before mounting. Mosaic images of each well were taken with SP8 (Leica) microscope. Analysis of the distance covered by cells labeled with DAPI and axons labeled with beta-III-tubulin was done in a semi-automatic mode using a custom ImageJ macro.

40To perform the time-lapse analysis of cortical neurons migrating on surfaces coated with FC (control), murine Lphn1 (Lec-Olf) or Lphn1 (Lec-Olf) TL-FL, cortical explants from E15.5 mouse embryos were cultured on 60-mm dishes that were coated for 30 min with 50 μg/ml of each protein in PBS. After 30 min, the dish was washed with PBS and coated with 20 μg/ml Laminin in PBS for at least 2 hours. The dish was next washed with PBS and cortical explants (E15.5) were cultured. After 2 hours in culture, samples were imaged with a Zeiss Axiovert 200M microscope equipped with a temperature-controlled carbon dioxide incubation chamber set to 37°C, 65% humidity and 5% CO2. Illumination was provided by an X-Cite lamp (series 120, Lumen Dynamics Group), and images were recorded by a Coolsnap HQ camera (Photometrics). Sequential images were acquired every 10 min. Analysis and tracking of all the neurons exiting the explant was carried out in Imaris (Bitplane) using the automatic tracking module.

41For time-lapse analysis of GFP+ neurons and axons challenged with alternate Lphn1(Lec-Olf) or Lphn1 (Lec-Olf) TL-FL stripes, E13.5 mouse embryos were electroporated with PCAGIG plasmid and cortical explants (E15.5) were placed on stripes as described above. After 2 hours in culture, cortical explants were imaged with a Zeiss Axiovert 200M microscope equipped with a temperature-controlled incubation chamber as described above. Sequential images were acquired every 10 min. Analysis was carried out using ImageJ, all contact events taking place in less than 2 frames were considered as repulsive events.

42Time-lapse experiments of GFP+ neurons on nanofibers started after 2 hours of culturing E15.5 cortical explants, that had previously been electroporated at E13.5. Samples were imaged with a Zeiss Axiovert 200M microscope equipped with a temperature-controlled incubation chamber as described above. Sequential images were acquired every 10 min.

43Cortical neurons (E15.5) were cultured on coverslips (13mm diameter, 24 well-plate, #1.5) coated with poly-D-lysine (Sigma, 0.5mg/ml) in Neurobasal medium (without phenol red for super resolution microscopy) supplemented with B27 (GIBCO Cat#17504044). After 2DIV, 0.2μg of Lphn1(Lec-Olf) or double mutant (Lec-Olf)TL-FL protein clustered with anti-his Alexa 488 antibody (Thermofisher cat#MA1-21315-A488, ratio 4:1 protein:Ig) was added to cortical neurons for 20 min at room temperature (RT). Neurons were washed twice with PBS and incubated with pre-clustered FLRT3 antibody (goat Ig, R&D, AF2795) with anti-goat Alexa-594 (Jackson, ImmunoResearch, Cat#705-585-003, ratio 4:1) and N-terminal Ten2 antibody (sheep, R&D, AF4578) with anti-sheep Alexa-647 (Jackson, ratio 4:1) for 20 min at RT. Neurons where washed twice with PBS and fixed for 20 min with 4% PFA at RT. Neurons were washed with ammonium chloride in PBS (50mM) for 5 min to quench the excess of PFA, followed by two washes with PBS and mounted with Immu-mount (Thermofisher, cat#10622689) for super resolution. In case of confocal microscopy, nuclei were stained with DAPI and samples were mounted with Dako (Agilent, cat#S3023). Confocal acquisition was done using a SP8 laser scanning confocal spectral microscope (Leica Microsystems Heidelberg). Images were taken using a 63 × objective with a 2 × digital zoom and 1 Airy disk pinhole. Co-localization was quantified using the co-localization plugin of Image (version 1.51p).

44Neurons were prepared as described above, and mounted in Immu-Mount (Thermo Fisher Scientific) for Single Molecule Localization Microscopy (SMLM) imaging on a Leica SR GSD system. The following settings where used for SMLM data acquisition: 642 nm laser at 40% of 500 mW output power for Alexa647, 560 nm laser at 100% of 500 mW output power for Alexa594 and 488 nm laser at 70% of 300 mW output power for Alexa488. For Alexa488 and Alexa647 low to moderate laser intensities of 405 nm were used for “back-pumping” to enhance single molecule blinking. For the position determination of the single fluorescent molecules we used a maximum-likelihood-based algorithm with a sliding window for background subtraction. SMLM images where reconstructed based on the localization accuracies of the individual single molecule positions.

45RNaseq data were accessed from the published NCBI Gene Expression Omnibus with accession numbers GEO:GSE10881 and GSE65000. Values for specific genes from individual cells were averaged and displayed as whisker plots.

46Embryonic brains were fixed in 4% PFA over-night and sectioned. 10 μm Cryo-sections were pre-treated using the RNAscope Universal Pretreatment Kit (Advanced Cell Diagnostics, Cat#322380). RNA In Situ Hybridizations (ISH) were performed using the RNAscope Fluorescent Multiplex Reagent Kit (Advanced Cell Diagnostics, Cat#320850) according to manufacturer’s instructions. The target genes (mouse Lphn1-3 and Ten1-4) are also listed in the Key Resources Table. Following ISH, sections were immunostained using mouse anti-Pvim 1/300 (Abcam, Cat#ab20346) and rat anti-Ctip2 1/600 (Abcam, Cat#ab123449) in combination with the Alexa Fluor 488 secondary antibodies 1/200 (Jackson Immunoresearch). Images were acquired using a Leica TCS SP8 confocal laser scanning microscope and processed with ImageJ software.

47Statistical analyses of were performed using GraphPad Prism, employing a two-tailed unpaired Student’s t test when comparing two groups (Figures S2G, 3C, 3F, 3I, S3H, S3L, S3P, and S3R) or one-way ANOVA test with Tukey’s post hoc analysis when comparing multiple groups (Figures 2E, 2F, 2I, 4B, 4H, S4B, S4D, 5B, 6D, 6F, 6G, S6E, 7C, 7E, 7G, S7B, and S7D) where P values represent ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001 and ∗∗∗∗p ≤ 0.0001. All data are presented as the mean ± s.e.m, whisker plots or dot plots. All sample sizes and definitions are provided in the figure legends.

48The crystallography data and models generated during this study are available for download in the Protein Data Bank (PDB accession numbers 6SKE, 6SKA). The code generated to analyze molecular dynamic simulations is freely available from https://github.com/MChavent/Hbond-analysis, or available upon reasonable request.