News & Views | Published:

Ion channels

How to be desensitized

The various functional states of glutamate receptors control much of the brain's neuronal activity. Our understanding of how one of those states — desensitization — occurs has taken a leap forward.

Neurons communicate with each other through synapses, which operate in the following way. On one side of the synapse, the incoming neuron converts electrical activity, encoded in action potentials, into a chemical signal by releasing a neurotransmitter. The membrane of the receiving, postsynaptic, neuron is packed with ion channels, which transduce neurotransmitter binding into complex conformational changes that induce channel opening and closing, a process known as gating. Channel gating translates the concentration profile of neurotransmitter in the synapse into a defined time course of ion flow across the postsynaptic neuronal membrane, changing the membrane potential accordingly and, usually, resulting in onward transmission of the presynaptic action potential. The most important class of neurotransmitter-gated ion channels conveying excitatory signals in the brain are the AMPA-type glutamate receptors. On page 245 of this issue, Sun et al.1 provide us with exciting snapshots of the gating process involved.

Like many other ion channels, AMPA receptors are doughnut-shaped structures, formed by the assembly of different members of the family of AMPA-receptor subunits2, that span the insulating neuronal membrane, with an ion-conducting water-filled pore down the centre. The pore is controlled by a gate, which 'opens' to permit ion flow only when the neurotransmitter — glutamate — binds to the receptor. However, AMPA receptors share with other ion channels a property known as 'desensitization', which limits the effective ion flow to the few milliseconds following binding of glutamate. This is because the channel tends to close even though neurotransmitter is still bound to it. This process has a considerable influence on how AMPA receptors encode information, as the degree of desensitization is a major factor in shaping the time course of postsynaptic responses3. Moreover, the degree of desensitization of the AMPA receptor is under exquisite control in various ways, such as regulated gene expression, RNA editing, alternative splicing of receptor-encoding RNAs, and the assembly of receptors from different subunits2. Despite the great importance of desensitization in brain function, however, little is known about how AMPA-receptor subunits coordinate gate opening and desensitization.

The extracellular glutamate-binding portion of each subunit comprises two protein domains. Earlier studies4 provided the crystal structure of these domains, which can best be described as resembling a clam shell. On binding glutamate, the cleft between the two shells (the two domains) narrows by the movement of the second domain towards the first5. This may directly trigger the opening of the gate, as amino-acid residues that link domain 2 to the plasma membrane are pulled away from the pore-forming region of the receptor. Because AMPA receptors are tetramers, and because all subunits are able to bind glutamate, they may unite in a stepwise fashion in the attempt to open the gate6. An unusual structural feature is that individual glutamate-binding pockets of AMPA receptors pair up to form dimers5. The dimer interface, which is formed by a twofold symmetric assembly through an interaction at the domain 1 interfaces, is packed with residues that change receptor desensitization dramatically if structurally modified3, indicating that the dimer interface may be involved in desensitization.

Sun et al.1 have used a combination of techniques in their study, and the results, taken together, are a major step forward in our understanding of how the AMPA receptor gates. First, they show that the stability — or rather lack of stability — of the dimer complex seems to be crucial for the desensitization process. They took the water-soluble glutamate-binding portions of the receptors (which they otherwise used to obtain the crystals) and measured their tendency to dimerize by sedimentation assay. When they compared the affinities with functional measurements of desensitization in the intact receptor, they found that a mutation that blocks desensitization7 increases the likelihood of dimerization of individual subunits by a factor of a hundred thousand. Similarly, dimerization of normal desensitizing receptors is greatly increased by the potent blocker of desensitization, cyclothiazide. Conversely, a mutation that accelerates desensitization reduces dimer affinity to a degree that it almost escapes detection. Comparison of dimer affinity and desensitization revealed an almost perfect inverse correlation between the two.

Next, with the aim of revealing the structural basis of the dimerization process, Sun et al. compared the crystal structures of mutant receptors that had dramatically different desensitization properties. The increased dimer affinity of the non-desensitizing mutant can be nicely explained by the formation of a strong hydrophobic pocket in the dimer interface. An equivalent effect can be seen in crystals of normal — wild-type — desensitizing receptors in the presence of cyclothiazide.

So what does the desensitization process look like? The authors' view of events is shown in Fig. 5 of the paper (page 251) and in Fig. 1 here. Activation of the receptor by glutamate, and subsequent movement of domain 2, produces tension on the connecting portion between the extracellular binding pocket and the receptor's transmembrane domains. This leads to conformational changes within the membrane, forcing the gate to open. But in consequence there is an equivalent strain on the glutamate-binding pocket, facilitating an ensuing conformational rearrangement to escape tension. This causes rearrangement at the dimer interface and movement of the receptor into the desensitized state.

Figure 1: Activation and desensitization of AMPA-type glutamate receptors.

a,The glutamate-binding domains are organized as a 'dimer of dimers', with the red and blue glutamate-binding cores forming one dimer and the purple and green subunits forming the other. The two dimers sit side-by-side so that the inter-dimer local twofold axis is in the centre of the four subunits and coincides with the fourfold axis of the ion channel. b, In the scheme of events proposed by Sun et al.1, binding of glutamate (yellow particles) induces domain closure at the glutamate-binding core: movement of domain 2 (D2) towards domain 1 (D1) causes conformational changes at the channel gate and opens the channel. c, Reorientation of the dimer interface decouples glutamate-binding-induced domain closure from the channel gate, and leads to receptor desensitization. (Modified from a graphic provided by Y. Sun.)

In receptors that do not desensitize, the strengthened dimer interface prevents rearrangement within the ligand-binding pocket. The linker between binding pocket and gate remains under tension and leaves the pore open as long as glutamate is bound. It is therefore no surprise that the glutamate-binding domain of the wild-type receptor and that of the non-desensitizing mutant both crystallize into the same 'open-channel' state, because the link, and therefore the tension force, from the membrane is missing.

Finally, Sun et al. surprise us with crystal structures of a mutant receptor that was designed to disrupt the dimer interface. The accelerated desensitization properties of the intact mutant receptor nicely confirm the importance of dimer affinity in desensitization. As expected, the structure does not dimerize at the usual interface. But it does reveal another, previously unknown, subunit interaction — hinting at a possible way in which pairs of dimers can assemble into a tetramer. In this interaction, domain 1 connects with a different region (lateral to that involved in dimerization) of domain 2 of a neighbouring receptor. If this interaction indeed takes place in the assembled receptor complex, the tetrameric structure represents a pair of dimers that are shifted parallel to one another.

This would be an unprecedented form of channel subunit assembly — one in which the dimers facing each other follow a different symmetry from the assumed (though not proven) fourfold symmetry of the receptor in the transmembrane channel pore. This arrangement may, however, provide a positional clue to how different isoforms of AMPA-receptor subunits assemble into a complex in vivo8, as chemically equivalent subunits may prefer to assume positions of equivalent symmetry, namely diagonally across from each other. It is yet not clear how these arrangements would affect receptor gating. But it is clear that glutamate receptors are likely to remain a rich source of surprises.


  1. 1

    Sun, Y. et al. Nature 417, 245–253 (2002).

  2. 2

    Dingledine, R., Borges, K., Bowie, D. & Traynelis, S. F. Pharmacol. Rev. 51, 7–61 (1999).

  3. 3

    Jones, M. V. & Westbrook, G. L. Trends Neurosci. 19, 96–101 (1996).

  4. 4

    Armstrong, N., Sun, Y., Chen, G. Q. & Gouaux, E. Nature 395, 913–917 (1998).

  5. 5

    Armstrong, N. & Gouaux, E. Neuron 28, 165–181 (2000).

  6. 6

    Rosenmund, C., Stern-Bach, Y. & Stevens, C. F. Science 280, 1596–1599 (1998).

  7. 7

    Stern-Bach, Y., Russo, S., Neuman, M. & Rosenmund, C. Neuron 21, 907–918 (1998).

  8. 8

    Mansour, M., Nagarajan, N., Nehring, R. B., Clements, J. D. & Rosenmund, C. Neuron 32, 841–853 (2001).

Download references

Author information

Correspondence to Christian Rosenmund.

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.