NMDA receptors are crucial in the workings of the brain and in its disorders. Two structures of almost complete receptors reveal the intricate complexity of these large, multi-domain molecular machines. See Article p.191
Since their discovery more than 30 years ago1,2,3, N-methyl-D-aspartate receptors (NMDARs) have fascinated neuroscientists. They are key mediators of synaptic plasticity, the cellular mechanism mediating information storage in the brain. Moreover, dysfunctions of NMDARs are implicated in various neuropsychiatric disorders, from schizophrenia to mental retardation and epilepsy, making them targets of therapeutic interest4. Two studies — one from Karakas and Furukawa5 published in Science, and another by Lee and colleagues6 on page 191 of this issue — present the long-awaited molecular structure of an NMDAR subtype, the GluN1–GluN2B receptor, from rats and frogs, respectively. These atomic maps offer unprecedented views and a wealth of information about a key brain receptor.
Together with the AMPA receptors (AMPARs) and kainate receptors, the NMDARs belong to the superfamily of ionotropic glutamate receptors (iGluRs), which mediate excitatory communication between neurons in the central nervous system. The iGluRs are massive protein complexes embedded in the neuronal cell membrane. Each complex contains more than 3,400 amino-acid residues, and is composed of four subunits. Each subunit has a typical modular architecture made up of a large extracellular amino-terminal domain (ATD) that participates in subtype-specific receptor assembly and modulation; a ligand-binding domain (LBD) that binds receptor-activating agonist molecules; a transmembrane domain (TMD) that forms an ion-channel pore; and a cytoplasmic carboxy-terminal domain (CTD) involved in receptor trafficking and coupling to intracellular signalling molecules.
However, NMDARs differ from other iGluRs in several respects. First, their transmembrane pore is highly permeable to calcium ions and is blocked by magnesium ions — features that are essential for triggering synaptic plasticity. Second, their LBDs require two different agonists for activation. And finally, their ATDs form a major regulatory region that harbours several binding sites for subunit-specific modulators4.
In a major feat, the first structure of a full-length iGluR was solved in 2009 — that of the AMPA GluA2 receptor7, which has four identical subunits. By contrast, structures of only bits of NMDARs have been solved8,9, and a complete structure showing ATDs linked to the gating core region (which forms from the LBDs and the TMD) has eluded all efforts. Solving the X-ray structure of an NMDAR is particularly challenging because it is formed from two types of subunit (two GluN1 and two GluN2 subunits) and is predicted to have a highly mobile ATD10,11,12.
Both Lee et al. and Karakas and Furukawa raced to solve the structure of GluN1–GluN2B using an approach similar to that used to solve the GluA2 receptor. This involved introducing a flurry of mutations at strategic locations in the receptor to trim flexible regions, including the entire CTD and some loops; reduce surface entropy; remove potential sites of sugar attachment and protein degradation; and eliminate reactive cysteine amino-acid residues, to prevent them from crosslinking. Each group got stuck because their protein crystals diffracted X-rays only to low resolution, presumably because of the flexibility of the molecules. To remove this barrier, both groups crosslinked the ATDs using disulphide bonds, restricting ATD conformational mobility. This improved the resolution of their structures to a reasonable 3.7–3.9 ångströms, but reduced the channel's ability to function quite drastically. The two teams eventually converged on remarkably similar structures, which probably represent a receptor trapped in an inactive or 'desensitized-like' state.
The GluN1–GluN2B receptor resembles a hot-air balloon, with the TMD at the 'bottom', the ATDs at the 'top' and the LBDs sandwiched in between (Fig. 1) — reminiscent of the layered organization of the GluA2 AMPAR. Notably, the GluN1 and GluN2B subunits alternate around the central pore, settling the lingering debate about subunit order. Also like GluA2, the four subunits assemble as a dimer-of-dimers and display 'domain swapping', such that pairing of domains between the ATD and LBD layers involves different subunits. Domain swapping thus seems to be a distinct, evolutionarily conserved structural feature of iGluRs that ensures subunit intertwining and overall cohesion of the large extracellular region.
The conservation of the tetrameric organization between NMDARs and AMPARs is somewhat surprising, considering the major differences in their ATD structures and in the ATD connections to the gating core. In the GluN1–GluN2B structure, the ATD heterodimers adopt the same peculiar twisted structure previously observed9 for isolated ATD domains. Moreover, the ATDs pack tightly against the LBDs, whereas ATD–LBD interactions in the GluA2 AMPAR are minimal. The extracellular region is therefore much more compact in NMDARs than in AMPARs; this characteristic is probably related to the unique role of NMDAR ATDs in modulating channel activity4.
Lee et al. provide the first structure of an almost complete iGluR pore, including the crucial pore loop — the short region that controls ion permeation. As expected13, the pore resembles an inverted potassium channel, and the symmetry of the pore loop indicates different contributions of the GluN1 and GluN2 subunits to permeation, as was previously predicted by electrophysiological studies14. Although neither research group provides a picture of a magnesium ion in its blocking site, Lee et al. do provide the first glimpse of channel-blocker sites, and Karakas and Furukawa highlight a potential calcium-ion storage site at the entrance of the pore. But more work is needed, including the determination of higher-resolution structures, to get a deeper understanding of the singular permeation properties of NMDARs.
What is the physiological relevance of the state captured in the structures, and what do the structures tell us about functional coupling between the ATD and gating-core regions? By assaying for domain proximity, Karakas and Furukawa convincingly show that the distinct inter-domain and inter-subunit interfaces observed in their crystal structure are present in an intact membrane-embedded receptor. When electrophysiological recordings of the crystallized receptors were made, however, the receptors were found to be functionally 'silent'. Functionality was recovered only when the authors broke the ATD crosslinks using reducing agents. This suggests that the conformational state of the ATDs in a functional receptor differs from that in the crystal. In a related result, Lee et al. observed that, in their low-resolution structure, the ATDs adopt a conformation in which the two ATD dimers split apart.
How conformational changes of the ATDs modulate receptors by influencing the gating core remains to be decrypted. In particular, the proposed10,11,12 large-scale motions of individual ATDs will have to be reconciled with the presence of extensive, presumably motion-restricting, interactions between the lower lobes of the ATDs and the upper lobes of the LBDs. Understanding the structural plasticity of the multi-domain receptor and the functional coupling between the various layers remains a major challenge. Nevertheless, the two new structures offer an exceptional framework for future structural, pharmacological and functional studies of NMDARs and of neurotransmission in general.