Structural basis for integration of GluD receptors within synaptic organizer complexes

See allHide authors and affiliations

Science  15 Jul 2016:
Vol. 353, Issue 6296, pp. 295-299
DOI: 10.1126/science.aae0104

Transmitting signals across the synapse

Glutamate receptors located on neuronal cells play a role in mediating electrical signals at excitatory synapses. These glutamatergic synapses are extremely important for nearly all cognitive functions. Elegheert et al. analyzed a complex that bridges the synapse, comprising β-neurexin 1, a cell adhesion molecule on the surface of presynaptic axons; cerebellin 1, a synaptic organizer; and the postsynaptic glutamate receptor GluD2. The structural and functional analysis provides insight into the mechanism of synaptic signaling.

Science, this issue p. 295


Ionotropic glutamate receptor (iGluR) family members are integrated into supramolecular complexes that modulate their location and function at excitatory synapses. However, a lack of structural information beyond isolated receptors or fragments thereof currently limits the mechanistic understanding of physiological iGluR signaling. Here, we report structural and functional analyses of the prototypical molecular bridge linking postsynaptic iGluR δ2 (GluD2) and presynaptic β-neurexin 1 (β-NRX1) via Cbln1, a C1q-like synaptic organizer. We show how Cbln1 hexamers “anchor” GluD2 amino-terminal domain dimers to monomeric β-NRX1. This arrangement promotes synaptogenesis and is essential for d-serine–dependent GluD2 signaling in vivo, which underlies long-term depression of cerebellar parallel fiber–Purkinje cell (PF-PC) synapses and motor coordination in developing mice. These results lead to a model where protein and small-molecule ligands synergistically control synaptic iGluR function.

Excitatory neurotransmission in the vertebrate central nervous system is largely mediated by the ionotropic glutamate receptor (iGluR) family members, classified as AMPA (GluA1-4), N-methyl-d-aspartate (NMDA) (GluN1, GluN2A-D, GluN3A-B), kainate (GluK1-5), and delta (GluD1-2) subtypes (1). All iGluRs are assembled from four modular subunits, each displaying extracellular amino-terminal and ligand-binding domains (ATD and LBD), a transmembrane domain (TMD) lining a central ion channel pore, and a cytoplasmic carboxy-terminal domain (CTD) (26). Binding of agonist molecules to the LBDs of GluA, GluN, and GluK receptors [typically glutamate, but also glycine or d-serine d-Ser for GluN subtypes] drives opening of the cation-conductive ion channel and neuronal membrane depolarization (1). Furthermore, all iGluRs appear to initiate postsynaptic signaling through nonionotropic mechanisms, a process better understood for NMDA and delta receptor subtypes (715).

In addition to their signaling function, mediated by excitatory amino acids, iGluRs are implicated in synaptogenesis (1620). The membrane-distal position of iGluR ATDs makes them readily accessible to other proteins populating the synaptic cleft. In this context, the GluD subfamily of iGluRs are the best studied. Cbln1, a soluble synaptic organizer molecule belonging to the C1q–tumor necrosis factor α (C1q-TNFα) superfamily (21), directly binds the GluD1 and GluD2 ATDs (16, 18, 22, 23). Cbln1 also interacts with presynaptic membrane-tethered neurexins (NRXs) (18) and establishes “molecular bridges” that span the cleft to facilitate bi-directional synaptic differentiation (16, 18). Despite their importance, the architecture of supramolecular GluD assemblies and the mechanisms by which they support integration of the dual synaptogenic and metabotropic signaling functions have remained unknown. We sought to address these questions by structurally investigating the β-NRX1–Cbln1–GluD2 transsynaptic triad.

As a first step toward solving a Cbln-GluD complex, we solved high-resolution crystal structures of (i) free human GluD2 (GluD2ATD, 1.75 Å) and mouse GluD1 (GluD1ATD, 2.30 Å) ATD dimers and (ii) free, nearly full-length human Cbln1 (Cbln1ΔVRSG, 2.80 and 7.00 Å) and the obligate Cbln1 globular domain trimer (with a C1q-like fold, Cbln1C1q, 2.35 Å) (figs. S1 to S4 and table S1) (24). Using surface plasmon resonance (SPR), we found that avidity governs the full-length Cbln1-GluD2 (Cbln1FL-GluD2FL) complex formation, which results in a nanomolar-range apparent affinity (KD,app of ~125 nM) (fig. S5) (24). Molecular dissection of the complex into individual components revealed that the Cbln1 C1q-like trimer is the minimal unit needed for interaction with GluD2ATD, with an affinity in the high micromolar range (figs. S5 and S6) (24). Attempts at cocrystallizing a Cbln1-GluD2ATD complex were hampered by the propensity of each component to crystallize separately. We consequently designed a construct that combines a fused Cbln1C1q trimer with GluD2ATD into one continuous polypeptide chain, linked by a 30-residue, flexible Gly-Gly-Ser [(G2S)10] spacer (Fig. 1A and figs. S7 and S8) (24, 25). We validated the mass and monodispersity of the Cbln1C1q-GluD2ATD chimera using multiangle light scattering (MALS) and negative-staining single-particle electron microscopy (EM) (fig. S8) (24).

Fig. 1 Architecture of the Cbln1C1q-GluD2ATD binary complex.

(A) Schematic representation of the chimeric Cbln1C1q and Cbln1C1q-GluD2ATD constructs. (B) “Front” and “side” view of the Cbln1C1q-GluD2ATD complex. The inward tilted orientation of the Cbln1 C1q domains suggests the position of the Cbln1 CRR, as visible in EM class averages and represented by a dotted ellipse here. GluD2 α helix 6′ and flap and cleft loops are highlighted. Disulfide bridges are shown as yellow spheres. (C) Symmetry-breaking in the Cbln1C1q interface. The total buried interaction surface is shown in a 90° rotated open-book view. (D) Selected negative-stain EM class averages of Cbln1FL illustrate its dimer-of-trimers arrangement. Yellow arrows indicate the suggested position of the CRR that links both C1q trimers. Scale bar, 10 nm.

We determined the crystal structure of the Cbln1C1q-GluD2ATD complex at 3.10 Å (Fig. 1B, fig. S8, and table S1). Cbln1C1q sits on top of the membrane-distal face of the GluD2ATD in an inward tilted orientation and breaks its three-fold symmetry to engage an ATD monomer with a total buried surface area (BSA) of 873 Å2 (Fig. 1C). The arrangement of both Cbln1C1q trimers suggests the position of the putative Cbln1 N-terminal “cysteine-rich region” (CRR, not present in the chimeric construct) that links two Cbln1C1q trimers into the hexameric Cbln1FL (Fig. 1B and fig. S4) (26). The distance between the calculated centers of mass of the trimers is 70 Å, in good agreement with the corresponding distances in two crystal structures of free Cbln1ΔVRSG (77 and 79 Å, respectively) (fig. S9). The tilted versus linear arrangement of the C1q trimers in the three structures suggests their intrinsic hinge movement relative to the CRR (fig. S9), consistent with the single-particle EM analysis (Fig. 1D and fig. S10). The Cbln1C1q-GluD2ATD structure, however, indicates that binding of GluD2 constrains the Cbln1C1q domain orientation. Single-particle EM class averages of free Cbln1FL suggested that the CRR is a globular structure linking two Cbln1 C1q trimers (Fig. 1D and fig. S10) (24).

In the Cbln1C1q-GluD2ATD crystal lattice, GluD2ATD forms the same N-shaped tetrameric “AB-CD” arrangement (Fig. 2A) previously observed in the full-length GluA2 and GluK2 structures (5, 27) and in structures of isolated GluA2 (28, 29) and GluK6 (30) ATDs. Cα-atom superposition of GluA2CRYST (5) and GluD2 ATD tetramers yields a root mean square deviation (RMSD) of 5.6  Å over 635 positions when aligning B-D dimers, and 7.3 Å over 695 positions when aligning A-C dimers (Fig. 2B). This AB-CD ATD arrangement is, as in GluA2CRYST, stabilized by two-fold symmetrical contacts between the D-B monomers (666 Å2 BSA) and consists of numerous putative salt bridges and hydrogen bonds, as well as potentially two calcium atoms (fig. S11) (24). Alignment of the Cbln1C1q-GluD2ATD complex to the full GluA2CRYST, using the B-D ATD dimer as reference, illustrates how one Cbln1 hexamer, binding to each ATD dimer, extends the outward-tapering, vertical Y-shape of the iGluR. The height of this arrangement (~17 nm) fits within the typical width of the PF-PC synaptic cleft (~20 nm) (Fig. 2C).

Fig. 2 Quaternary structure of the β-NRX1–Cbln1–GluD2 complex.

(A) “Top” view of the full Cbln1C1q-GluD2ATD dimer-of-dimers complex. The black ellipse, black arrows, and red arrows indicate the overall two-fold symmetry axis, the two-fold symmetry axes in the GluD2ATD dimers, and the three-fold symmetry axes in the Cbln1C1q trimers, respectively. The suggested position of the Cbln1 CRR is marked with dashed ovals. (B) Superposition of the GluD2 and GluA2 (PDB 3KG2) (5) N-shaped ATD layers using the B-D ATD dimers [view equivalent to (A)]. Centers of mass of GluD2ATD (black spheres) and GluA2ATD (red spheres) are connected to highlight overall similarity. (C) View along the overall two-fold axis of the Cbln1C1q-GluD2ATD complex aligned to Y-shaped GluA2CRYST using the B-D ATD dimers. (D) Selected negative-stain EM class averages of the β-NRX1(+4)–Cbln1FL complex. Yellow arrows indicate the suggested position of β-NRX1(+4). Scale bar, 10 nm. (E) Model of the synapse-spanning β-NRX1(+4)–Cbln1–GluD2 complex.

We then investigated how Cbln1 links GluD2 to presynaptic β-NRX1. The extracellular part of the single-span membrane-tethered β-NRX1 consists of a single LNS (laminin, neurexin, sex hormone–binding globulin) domain, LNS6, of known structure (31). Insertion of the 30-residue “spliced sequence #4” (SS4) into β-NRX1LNS6 via alternative splicing, to yield β-NRX1(+4), is required for Cbln1 binding (18). Using isothermal titration calorimetry (ITC), we measured that one hexameric Cbln1 and one monomeric β-NRX1(+4) bind with high affinity (KD = 43.5 ± 4.4 nM) and that the Cbln1 CRR and β-NRX1 SS4 mediate all binding (fig. S12). Notably, the 1-to-1 stoichiometry indicates that the two-fold Cbln1 symmetry is broken in β-NRX1(+4)–Cbln1FL. Single-particle negative-stain EM class averages of β-NRX1(+4)–Cbln1FL confirm this stoichiometry and confirm that the Cbln1 CRR is the β-NRX1(+4) binding platform (Fig. 2D and fig. S10) (24). Together, our structural analyses allow us to propose an overall model for the β-NRX1–Cbln1–GluD2 triad that features symmetry mismatches in both binary complex interfaces (Fig. 2E).

Cbln1C1q binds GluD2ATD at the membrane-distal ends of α helices 1, 2, 3, 9 and 10 of the R1 lobe and at the “flap” loop, an extended, structurally conserved segment that links α helices 9 and 10 and folds back onto the top of the ATD. Cbln1 interface residues are contributed by loops AA′ (Ser69-Ser75), CD (Tyr122-Thr126), EF (Gly144-Arg150), and GH (Gly174-Lys181) from one C1q subunit (Fig. 3A). Cbln1 loops EF and CD rearrange upon binding the relatively flat top of GluD2ATD, which remains largely unchanged between free and Cbln1-bound states (overall Cα-atom RMSD of 0.5 Å over 336 positions) (Fig. 3A).

Fig. 3 Details and structure-guided mutagenesis of the Cbln1-GluD2 interface.

(A) Superposition of free and bound Cbln1C1q and GluD2ATD. (B) GluD2 alanine-scanning mutagenesis; SPR response levels are color-annotated as a heat map onto the GluD2 ATD structure. The mutated interface is outlined in black, and the interaction hotspot is outlined in red. (C) GluD2 alanine-scanning mutagenesis using Cbln1FL and monomerized GluD2ATD. The chart shows absolute SPR responses after stimulation with 100 μM Cbln1, relative to wild-type GluD2. (D) Quantification of hemi-synapse formation by GCs and HEK 293T cells expressing structure-guided GluD2 mutants. Syn, synaptophysin. Data represent means ± SEM. ****P < 0.0001 (Kruskal-Wallis and Steel-Dwass test). (E) Quantification of contacted synapses between PFs and PCs expressing structure-guided GluD2 mutants in Grid2-null (GluD2-deficient) cerebella. Data represent means ± SEM. ****P < 0.0001; n.s., not significant (Kruskal-Wallis and Steel-Dwass test).

Cbln1 Tyr122 forms the central interaction hotspot; it is oriented by GluD2 Glu61 and Thr60 and buried in a small hydrophobic pocket formed by GluD2 Ile26, Ile27, Leu58, Leu342, and Trp347 (fig. S13). At the periphery, Asp24, the first residue of the mature GluD2 chain, is locked into an extended putative hydrogen-bonding network by Cbln1 residues Thr70, Asn71, and His72 of loop AA′, and Lys181 of loop GH. Cbln1 Arg150 and Asp147 on loop EF engage in putative charged interactions with GluD2 Glu61 and Arg345, respectively (fig. S13). Sequence conservation analysis indicates that the flap loop interface residues of GluD1/2ATD and the loop-based interface residues in the Cbln family (Cbln1 to Cbln4) are highly conserved in vertebrates (fig. S14).

We performed single-position alanine-scanning mutagenesis on 15 contiguous GluD2 interface residues to validate the Cbln1C1q-GluD2ATD binding mode. These mutations containing an amino acid substitution (25)—D24A, S25A, I26A, E61A, L342A, R345A, H348A, or S352A—cluster into the center of the observed interaction interface, and they showed reduced binding to Cbln1 in an avidity-enhanced SPR setup. Seven peripheral mutations maintained binding: T60A, E343A, D344A, K346A, S349A, M350A, and Q364A (Fig. 3, B and C, and fig. S15). Furthermore, the Cbln1FL, Cbln1C34S,C38S (lacking the CRR), and Cbln1C1q variants containing the Y122A, R124A, and D147A interface mutations lost all binding to GluD2ATD (fig. S16).

The expression pattern of GluD2, specifically confined to PF-PC synapses in the cerebellar cortex (32), as well as the availability of well-characterized Grid2-null (GluD2-deficient) and Cbln1-null mice, offered us the opportunity to validate and mechanistically interrogate structural models in a series of functional assays, from cell culture to in vivo. Guided by the Cbln1C1q-GluD2ATD structure, we designed a number of GluD2 mutants that target distinct structural features of the receptor to gauge their effect on the two well-established major functions of GluD2 signaling: (i)PF-PC synapse formation and (ii) induction of long-term depression (LTD) of synaptic transmission (8, 32). GluD2D24A,I26A,E61A,R345A combines four Ala mutations that abolish Cbln1 binding (Fig. 3B and fig. S17). GluD2E343A,K346A,S349A,M350A combines four Ala mutations that maintain Cbln1 binding but are immediately adjacent to the binding hotspot (Fig. 3B and fig. S17). GluD2F76D contains monomerized ATDs (fig. S6); this mutation was specifically designed to disrupt the N-shaped ATD layer, native Cbln1FL-GluD2ATD–binding geometry and overall Y-shape of GluD2 (Fig. 2C), while maintaining the Cbln1C1q-binding platform (Fig. 3A). Finally, GluD2ATD-LBD_GLYCAN_WEDGE contains a 10-residue glycosylated linker [ELSNGTDGAS in single-letter amino acid code (25)] inserted between the ATD and LBD layers in order to space them apart and disrupt potential mechanical ATD-LBD coupling (fig. S17) (24).

The effect of the glycan wedge (GW) on ATD-LBD coupling was tested by introducing the Ala654Thr mutation [“Lurcher” (Lc), located in the transmembrane (TM) helix M3] into GluD2 to render the channel constitutively open (fig. S18). Application of the LBD agonist d-Ser inhibited the Lc current in a dose-dependent manner, as expected (33). However, the half-maximal inhibitory concentration (IC50) in GluD2GW/Lc was ~1.5 times that compared with wild-type (WT) GluD2 carrying the LC mutation (GluD2WT/Lc), which indicated that separation of ATD and LBD by the glycan wedge impaired the ability of the LBD to induce pore closure (fig. S18).

We set up a heterologous hemi-synapse formation assay in which human embryonic kidney 293T (HEK 293T) cells expressing these mutants were cocultured with isolated wild-type mouse cerebellar granule cells (GCs) (Fig. 3D and fig. S17). Consistent with our mutagenesis data, cells expressing GluD2D24A,I26A,E61A,R345A accumulated no GC axon terminals, whereas cells expressing GluD2E343A,K346A,S349A,M350A did so normally. Cells expressing GluD2F76D showed intermediate GC axon terminal accumulation, which suggested that a disrupted ATD dimerization interface still allows synapse-spanning interactions to a certain extent. The GluD2ATD-LBD_GLYCAN_WEDGE mutant showed normal GC axon terminal accumulation, consistent with the notions that the Cbln1C1q-GluD2ATD complex interface and the overall receptor geometry were preserved.

We used recombinant Sindbis viruses to back-express either GluD2WT or GluD2 mutants in the cerebellum of adult Grid2-null mice older than postnatal day 30 (>P30). Grid2-null mice typically display a marked (~40%) reduction of PF-PC synapse number (32). Using immunohistochemical analysis of ultrathin sections with gold-conjugated antibodies against GluD2 (anti-GluD2), we detected that all mutants reached the PC postsynaptic density (PSD) (fig. S19). We counted the number of PF-PC synapses in the EM micrographs; the proportion of contacted synapses for GluD2E343A,K346A,S349A,M350A and GluD2ATD-LBD_GLYCAN_WEDGE was comparable to the proportion for GluD2WT. GluD2D24A,I26A,E61A,R345A and GluD2F76D, however, failed to robustly induce PF-PC contacts (Fig. 3E and fig. S19). Thus, disruption of the Cbln1-GluD2 interface and binding geometry attenuates rapid induction of PF-PC synapses.

Cerebellar LTD is caused by activity-induced endocytosis of postsynaptic AMPA receptors in PC dendrites (34). GluD2 signaling via the CTD is absolutely required for LTD induction, independent of PF synapse formation; LTD is impaired even when the PF synapse number is restored by reintroduction of GluD2 lacking the CTD in Grid2-null mice (35). d-Ser binds and closes the LBD of GluD2 (fig. S18) (33) to enhance LTD induction via the CTD of GluD2 (8). Application of exogenous d-Ser (200 μM for 10 min) reduced PF-evoked excitatory postsynaptic currents (PF-EPSCs) in mature wild-type, but not in Cbln1-null PCs, in the presence of NMDA receptor blockers to prevent coactivation of NMDA receptors (fig. S20). Endogenous d-Ser is released at PF-PC synapses from neighboring Bergmann glia by burst stimulation (BS) of PFs in immature cerebellar slices (8). Indeed, conjunctive BS and direct PC depolarization (BS/ΔV) (Fig. 4A) induced a robust LTD in wild-type but not in Cbln1-null or Grid2-null PCs in the presence of NMDA receptor blockers (fig. S20). These results indicate that GluD2 requires the presence of Cbln1 in order to respond to exo- or endogenous d-Ser and trigger AMPA receptor endocytosis.

Fig. 4 Structure-guided GluD2 mutants affect cerebellar synaptic plasticity.

(A) Setup to induce LTD in immature PC dendrites. BG, Bergmann glia; Rec, recording electrode; BS, burst PF stimulation; ΔV, direct PC depolarization; NMDAR, NMDA receptor. (B) Averaged LTD data from immature Grid2-null (GluD2-deficient) PCs expressing GFP + GluD2 variants, after BS/ΔV (arrow). (Insets) PF-EPSCs at t = –1 min and t = 30 min time points relative to BS/ΔV application. Data represent means ± SEM. **P < 0.01; *P < 0.05; n.s., not significant (Kruskal-Wallis and Steel-Dwass test). NMDA receptor blockers are 100 μM d-AP5 plus 25 μM MK801. (C) Proposed key events leading to signal transmission in the β-NRX1(+4)–Cbln1–GluD2 triad. (I) β-NRX1–Cbln1 is a presynaptic anchor for GluD2. (II) Transsynaptic complex formation. GluD2 allows binding of two β-NRX1(+4)–Cbln1 complexes and is shown at full occupancy. (III) GluD2, Cbln1, and d-Ser cooperatively induce postsynaptic LTD.

We examined whether the structure-guided GluD2 mutants are able to support d-Ser–dependent LTD at PF-PC synapses. BS/ΔV induced LTD in immature Grid2-null PCs expressing GluD2WT and GluD2E343A,K346A,S349A,M350A but not GluD2D24A,I26A,E61A,R345A, GluD2F76D, or GluD2ATD-LBD_GLYCAN_WEDGE (Fig. 4B and fig. S21) (24). Thus, anchoring by Cbln1, stable ATD dimer formation, and ATD-LBD coupling are all required for GluD2 to mediate the d-Ser–dependent LTD signals.

To relate our findings to cerebellar function in vivo, we subjected immature Grid2-null mice virally expressing GluD2WT and the structure-guided GluD2 mutants to an accelerating (4 to 40 rpm. in 5 min) rotor-rod test. Conventional cerebellum-dependent learning tasks, such as eye blink-conditioning, cannot be used in immature mice, when d-Ser is still present. Motor coordination was recovered after expression of GluD2WT, GluD2E343A,K346A,S349A,M350A, and GluD2ATD-LBD_GLYCAN_WEDGE but not of GluD2D24A,I26A,E61A,R345A or GluD2F76D (fig. S22). These results reflect our PF-PC synapse formation data and suggest that the interface and geometry of the Cbln1-GluD2 complex we describe is crucial for restoration of PF-PC cerebellar circuitry and motor-related performance.

Our results provide a molecular framework for the large body of studies on the β-NRX1–Cbln1–GluD2 transsynaptic signaling system and suggest a three-step model for GluD2 signaling activation. Cbln1 is secreted from cerebellar GCs through as-yet-unidentified mechanisms and remains associated with β-NRX1 on the surface of GC axons (step I in Fig. 4C) (36). When cerebellar GC axons encounter PC dendritic spines, the β-NRX1–Cbln1 complex “hooks” GluD2, via avidity-enhanced Cbln1C1q-GluD2ATD interactions, in a transsynaptic complex (step II in Fig. 4C). The biological importance of the binding avidity most likely lies in improving the probability for molecular recruitment of GluD2 by β-NRX1–Cbln1 at PF-PC synapses, given the weak affinity between individual Cbln1 and GluD2 domains. It is, however, unclear whether GluD2 is fully occupied at any given time point. Finally, agonist binding triggers a conformational change of the GluD2 LBD that is transmitted to the transmembrane domain to initiate downstream signaling (9), which results in AMPA receptor endocytosis and LTD (8, 9) (step III in Fig. 4C).

Recent structures of isolated full-length GluA and GluK receptors in different functional states have established that iGluR gating is accompanied by complex ATD-LBD relative motions (27, 37, 38). Furthermore, desensitization of GluA receptors results in a marked conformational variability in the ATD layer (4, 27, 37). Similar motions might, in principle, be possible in GluD receptors, considering their overall structural and sequence homology to GluA and GluK family members. However, we propose that, in a transsynaptic context, anchoring of GluD to the β-NRX1(+4)–Cbln1 complex will limit or prevent large-scale motions of the ATD layer. As a result, the force generated by LBD closure, driving the overall receptor contraction, is likely to transfer predominantly toward the postsynaptic membrane. Because most, if not all, iGluR family members are anchored to synaptic cleft proteins via their ATDs (1620), it is conceivable that their range of motions may also differ from those currently described in isolated receptors (4, 27, 37). Such interactions will impact on both iGluR location and conformation in response to agonist binding. We propose that the concept illustrated here, where small molecule and protein ligands cooperate in order to modulate GluD2 signaling, is likely to be more generally applicable to neurotransmitter receptors.

Correction (29 July 2016): Due to an error by Science staff, several author changes were not made before publication. These changes included the removal of a duplicate reference and a correction to a sentence describing the half-maximal inhibitory concentration (IC50) in GluD2GW/Lc compared with the wild type carrying the Lc mutation (GluD2WT/Lc). The HTML and PDF have been updated to reflect these changes.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S22

Table S1

References (39103)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
Acknowledgments: We thank staff at Diamond Light Source, T. Walter and K. Harlos for crystallization technical support, N. Scull for assistance with molecular biology, and E.Y. Jones and P. Miller for comments on the manuscript. This work was funded by the UK Medical Research Council (MRC) (G0700232 and L009609 to A.R.A. and MC_U105174197 to I.H.G.), the Japan Society for the Promotion of Science (15H05772 to M.Y. and 26117515, 26293042 to W.K.), the Core Research for Evolutional Science and Technology (CREST), Japan Science and Technology Agency (to M.Y.), the Takeda Science Foundation (W.K. and M.Y.), the Yamada Science Foundation (W.K.), the Human Frontier Science Program (RGP0065/2014 to M.Y. and A.R.A.), and the NIH (R01HD061543 to T.N.). The Wellcome Trust Centre for Human Genetics is supported by Wellcome Trust grant 090532/Z/09/Z. J.E. was supported by European Molecular Biology Organization (ALTF 1116-2012) and Marie-Curie (FP7-328531) fellowships, C.S. is a Cancer Research UK senior research fellow (C20724/A14414), and A.R.A. is an MRC senior research fellow (MR/L009609/1). Structure factors and coordinates of Cbln1C1q, Cbln1ΔVRSG crystal forms 1 and 2, GluD2ATD, GluD1ATD, and Cbln1C1q–GluD2ATD are deposited in the Protein Data Bank (PDB codes 5KC5, 5KC6, 5KC7, 5KC8, 5KC9 and 5KCA, respectively).

Stay Connected to Science

Navigate This Article