Latrophilin 1 and its endogenous ligand Lasso/teneurin-2 form a high-affinity transsynaptic receptor pair with signaling capabilities

Isolation of the LPH1 Ligand.

To create an affinity adsorbent, three soluble constructs based on the NTF of LPH1 were expressed in neuroblastoma (NB2a) cells (Fig. 1A): LPH-51, containing a V5 tag at the N terminus and the natural cleavage sequence at the C terminus; LPH-74, terminating exactly at the cleavage site; and LPH-62, with a V5 tag attached at the cleavage site but not cleaved in cells. α-LTX binding to these proteins demonstrated biological activity in only LPH-51 (Fig. 1B).

Purified LPH-51 was attached to an immobilized anti-V5 Ab and used for affinity chromatography of a detergent extract from rat brain. Bound proteins were desorbed using high pH, which also eluted LPH-51 but not the Ab. We used two control conditions: a V5 Ab/LPH-51 column not incubated with brain extract and a V5 Ab column (without LPH-51) incubated with the extract. A single brain protein of ∼275 kDa was eluted from the test column, but not from the control columns (Fig. 1C). To explore the nature of this protein, we used two additional V5 Ab/LPH-51 columns, one incubated with a detergent-free extract from brain and the other incubated with a detergent extract from rat liver. The 275-kDa protein did not bind to control columns (Fig. 1D).

Initial MS analysis failed to identify the LPH1-binding protein, but the affinity chromatography results (Fig. 1D) indicated that the isolated protein was brain-specific and membrane-bound. Thus, it was operationally termed Lasso (LPH1-associated synaptic surface organizer). For further analysis, we isolated a large amount of Lasso and elicited anti-Lasso Abs in mice.

These Abs (dmAbs) were used to test whether Lasso interacts with endogenous LPH1. We purified LPH1 from rat brain lysate by affinity chromatography on α-LTX or anti-LPH1 Ab and found that Lasso was isolated together with LPH1 in both cases, but not on a control column (Fig. 1E). Thus, in the brain, LPH1 and Lasso form strong complexes that survive stringent purification conditions.

We also used high-resolution tandem MS for de novo sequencing of Lasso and determined the sequences of nine tryptic peptides (Table S1). All of the peptides corresponded precisely to neurestin (26), also known as rat teneurin-2.

Based on the peptide sequences obtained, cDNA encoding full-size (FS) human Lasso was cloned (Fig. 2A). Lasso cDNA is 12,450 bp long and encodes a protein which has a calculated molecular mass of 292 kDa, no signal peptide, one predicted TMR, and is a splice variant of teneurin-2 (Fig. 2A).

Lasso/Teneurin-2 Is a Ligand of LPH1.

To examine the interaction between Lasso and LPH1, we expressed Lasso as membrane-bound (Lasso-A) or secreted (Lasso-B to -H) proteins (Fig. 2A). These constructs were produced in stably transfected NB2a cells (Lasso-A to -F), transiently transfected HEK293 cells (Lasso-H), or bacteria (Lasso-G expressed as a GST fusion).

We tested the truncated soluble proteins for LPH1 binding by a pull-down assay (Fig. 2B). Lasso-B, -D, and -E strongly bound LPH-51, whereas Lasso-C and -F showed no appreciable binding (Fig. 2B), suggesting that LPH1 binding localized to Lasso's C-terminal globular domain. To confirm this, we tested Lasso-H in a cell-binding assay and found that it interacted specifically with LPH1-FS-expressing cells (Fig. S1 C and G).

We also prepared truncated constructs of LPH1 (Fig. 2C). Compared with LPH-51, these constructs displayed no affinity for Lasso (Fig. 2D). This lack of binding was not due to misfolding of the LPH1 mutants, because constructs LPH-97–99 strongly bound α-LTX (Fig. 2E), indicating that at least these proteins folded correctly. As all truncations in LPH1 led to a loss of binding to Lasso, but not to α-LTX, and considering the large size of the globular domain of Lasso, the latter likely binds LPH1 at multiple points, which cooperate to provide a high-affinity interaction.

Which of the three LPH homologs expressed in mammalian brain (3) binds Lasso specifically? We incubated Lasso-D with rat brain lysate and immunoprecipitated it. Analysis of the eluate with Abs against each LPH homolog (36) identified LPH1 as the main ligand of Lasso-D (Fig. 2F). Compared with the brain content of each LPH homolog, the binding was ranked as follows: LPH1, 100% ± 10%; LPH2, 9% ± 8%; LPH3, 0.5% ± 0.1% (n = 3).

The dissociation constant, K_d, of the Lasso–LPH1 complexes was found to be 1.7 ± 0.2 nM for Lasso-D (Fig. 2G).

Head domains of teneurin-1, -3, and -4 (similar to Lasso-H) (Fig. S1A) were also expressed in HEK293 cells (Fig. S1H) and tested for LPH1 binding; however, unlike Lasso-D and -H (Fig. S1 B and C), they demonstrated no interaction (Fig. S1 D–G). This agrees with the MS data showing that only teneurin-2 was purified from brain on the LPH1 column (Table S1).

Lasso and LPH1 Interact Across Artificial and Synaptic Junctions.

We then verified that LPH1 and Lasso also interact on the cell surface. Lasso-D and LPH-51 were incubated with NB2a cells expressing LPH1-FS and Lasso-A, respectively. Immunostaining showed that each soluble protein bound only to those cells that expressed its respective pair (Fig. 3 A and B).

To test whether Lasso and LPH1 also interact across intercellular junctions, we cocultured cells expressing LPH1-FS with cells expressing Lasso-A. As a control, Lasso-A cells were cocultured with cells expressing another adhesion GPCR, EMR2. We found that in LPH1/Lasso-A cocultures, the cells clearly preferred heterophilic interactions, i.e. LPH1 cells tended to aggregate with Lasso cells, forming multiple contacts (Fig. 3 C and D and Fig. S2A). The staining for each protein overlapped at cell contacts, to which both proteins were recruited (Fig. 3D and Fig. S2A). These effects were not observed in control cocultures (Fig. 3C). Quantification of colocalized immunostaining (Fig. 3E) showed that LPH1 and Lasso interacted specifically across intercellular junctions.

Similar to the unrelated receptor EMR2 (Fig. 3C), LPH2-FS, a LPH1 homolog that poorly binds Lasso, also failed to mediate cell aggregation, as demonstrated by the number and size of cell clusters formed (Fig. 3F and Fig. S2B). Thus, the attraction between LPH1 and Lasso cells is caused by a strong interaction between these proteins.

Both LPH1 and Lasso are found exclusively in the brain (Fig. 4 A and B). As this pair mediates cell–cell interactions in culture, could it participate in synaptic contacts? Fractionation of rat brain membranes revealed that both proteins were largely absent from the cell bodies (P1), but concentrated in isolated synapses and synaptic plasma membranes (Fig. S3A).

Immunostaining of rat hippocampal cultures (Fig. 4C and Fig. S3) revealed that endogenous LPH1 was absent from cell bodies, but present in axons (Fig. S3B), and enriched in growth cones and synapses, colocalizing well with synapsin and partially with PSD-95 (Fig. 4C and Fig. S3 C–G). Lasso was expressed mainly on dendrites (Fig. S3H), concentrating in dendritic spines and large patches on dendritic shafts (Fig. 4C and Fig. S3 I–L) and colocalizing partially with synapsin and strongly with PSD-95 (Fig. 4C and Fig. S3 I–J). Costaining for LPH1 and Lasso revealed puncta of distinct sizes (smaller for LPH1 and larger for Lasso) that often colocalized in shaft synapses (Fig. 4C and Fig. S3 K and L) and almost always in varicosities formed at neurite intersections (Fig. S3 M and N). The degree of LPH1 and Lasso colocalization with synaptic markers (Fig. 4D) appeared to be consistent with the two proteins residing on opposite sides of the synapse.

The synaptic localization of LPH1 and Lasso was confirmed by immunoelectron microscopy. To identify LPH1, synaptosomes were incubated with V5-tagged A1 Ab and gold-conjugated anti-V5 Ab (Fig. 4 E and F and Fig. S4). To localize Lasso, we incubated synaptosomes with LPH-51, then stained them with gold-conjugated anti-V5 Ab. This approach was tested on hippocampal neurons (Fig. S4A), before being applied in immunoelectron microscopy of synaptosomes (Fig. 4G and Fig. S4). The distribution of gold particles (Fig. 4E) showed that LPH1 was present on nerve terminals near synaptic clefts, but not on dendritic spines, whereas Lasso was widely present on dendritic spines and in the cleft.

To quantify the distribution of Lasso and LPH1 at synaptic junctions, we isolated rat brain synapses, dissociated them into pre- and postsynaptic junctional membranes (37), and analyzed by immunoblotting using voltage-gated K⁺ channel (Kv1.2) and PSD-95 as pre- and postsynaptic markers, respectively. Further analysis of the membrane fractions (Fig. 4H) revealed that LPH1 was enriched in the presynaptic membranes, whereas Lasso was mostly postsynaptic (Fig. 4I). We verified our method using the known distribution of the NMDA-type glutamate receptor (Fig. 4I).

Taken together, these data indicate that LPH1 and Lasso are present in synaptic junctions and can form transsynaptic contacts.

LPH1 and Lasso in Chimerical Synapses.

By forming transsynaptic pairs, LPH1 and Lasso could play a role in synaptognenesis. This prompted us to test this pair in artificial synapse formation. We cocultured primary hippocampal neurons with HEK293 cells expressing Lasso-A or LPH1-FS and immunostained these cocultures for Lasso or LPH1 and for the neuron-specific PSD-95 or synapsin 1. These experiments showed that LPH1-FS on the surface of HEK cells was recruited to contacts with passing dendrites, which respectively accumulated PSD-95 at these sites (Fig. 5B). By contrast, Lasso-A on HEK cells did not accumulate at artificial synapses marked by postsynaptic PSD-95 (Fig. 5A). Reciprocally, LPH1-FS did not colocalize with synapsin-positive axonal varicosities surrounding LPH1-FS–expressing cells (Fig. S5B), while Lasso-A did (Fig. S5A). In addition, Lasso-A induced strong axonal attraction (Fig. S5A), diverting axons from their original growth directions and causing them to envelop the Lasso-A+ cells. LPH1-FS expressed on HEK cells showed much lower axon-attracting activity and no ability to divert axonal growth (Fig. S5B). These data indicate that while Lasso has a strong effect on axonal pathfinding and axon–target interaction, LPH1 is capable of inducing postsynaptic specializations.

C-Terminal Fragment of Lasso Induces LPH1 Signaling.

What is the consequence of the LPH1–Lasso interaction? Hippocampal neurons and heterologous LPH1-expressing cells respond to LTX^N4C by increasing cytosolic Ca²⁺ (5, 6). To assess the ability of Lasso to stimulate similar signaling, we used Lasso-G (Fig. 2A), which was able to bind NB2a cells expressing LPH1-FS (Fig. S6A). The cells were loaded with the fluorescence Ca²⁺ indicator Fluo-4 and continuously imaged. When LTX^N4C (positive control) was added in the absence of Ca²⁺, no signaling occurred, as expected from our previous experiments (5, 15); Lasso-G behaved similarly (Fig. S6B). However, when extracellular Ca²⁺ was added, the Ca²⁺ signal in the presence of LTX^N4C or Lasso-G was substantially stronger and longer-lasting than in negative control (P = 0.001) (Fig. S6 B and C). Thus, similar to LTX^N4C, the C-terminal fragment of Lasso can induce LPH1-mediated signaling to Ca²⁺ stores in LPH1-expressing cells.

We further tested the effect of Lasso-G on presynaptic Ca²⁺ dynamics in cultured hippocampal neurons. We simultaneously loaded a selected neuron with a fluorescent morphological tracer and Fluo-4 via a somatic patch-pipette in whole-cell mode (Fig. 6A; SI Methods). Using tracer fluorescence, we identified axons and presynaptic boutons (Fig. 6B) and measured presynaptic Ca²⁺ fluorescence before and after application of Lasso-G. Again, we found that, unlike the negative control, the C-terminal fragment of Lasso induced substantial increases in presynaptic Ca²⁺ concentration (Fig. 6 C and D).