Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/6207803

MoleculardeterminantsofAMPAreceptorsubunitassembly

ArticleinTrendsinNeurosciences·September2007

DOI:10.1016/j.tins.2007.06.005·Source:PubMed

CITATIONS

114

READS

82

3authors,including:

IngoGreger

UniversityofCambridge

59PUBLICATIONS1,535CITATIONS

SEEPROFILE

AndrewCharlesPenn

UniversityofSussex

19PUBLICATIONS575CITATIONS

SEEPROFILE

AllcontentfollowingthispagewasuploadedbyAndrewCharlesPennon01December2016.

Theuserhasrequestedenhancementofthedownloadedfile.Allin-textreferencesunderlinedinbluearelinkedtopublicationsonResearchGate,lettingyouaccessandreadthemimmediately.

Molecular determinants of AMPAreceptor subunit assemblyIngo H. Greger1, Edward B. Ziff2 and Andrew C. Penn1

1 MRC Laboratory of Molecular Biology, Neurobiology Division, Cambridge CB2 2QH, UK2 NYU School of Medicine, Department of Biochemistry, New York, NY 10016, USA

Review TRENDS in Neurosciences Vol.30 No.8

AMPA-type (a-amino-3-hydroxy-5-methyl-4-isoxazolepr-opionate) glutamate receptors (AMPARs) mediate post-synaptic depolarization and fast excitatory transmissionin the central nervous system. AMPARs are tetrameric ionchannels that assemble in the endoplasmic reticulum (ER)in a poorly understood process. The subunit compositiondetermines channel conductance properties and gatingkinetics, and also regulates vesicular traffic to and fromsynaptic sites, and is thus critical for synaptic functionand plasticity. The distribution of functionally differentAMPARs varies within and between neuronal circuits, andeven within individual neurons. In addition, synapsesemploy channels with specific subunit stoichiometries,depending on the type of input and the frequency ofstimulation. Taken together, it appears that assemblyis not simply a stochastic process. Recently, progresshas been made in understanding the molecular mechan-isms underlying subunit assembly and receptor biogen-esis in the ER. These processes ultimately determine thesize and shape of the postsynaptic response, and are thesubject of this review.

IntroductionIonotropic glutamate receptors (iGluRs) mediate themajority of fast excitatory synaptic transmission in vert-ebrate central nervous systems (CNSs) [1]. Three subfami-lies of glutamate-gated, cation-selective channels mediaterapidsignalling:AMPA,NMDA(N-methyl-D-aspartate)andkainate receptors. They play pivotal roles in synapse for-mation and maintenance, and in various forms of synapticplasticity [2,3]. In addition, dysfunction of these receptorsresults in diverse acute and chronic neurological disorders.iGluRs are widely distributed in the CNS, where theyfulfil different functions. AMPA-type glutamate receptors(AMPARs) mediate primary depolarization in glutamater-gic transmission, and their exceptionally fast kinetics con-veys ‘point-to-point’ signalling [2,4].

In contrast to the skeletal nicotinic acetylcholinereceptor, which is built in a predetermined fashion [5],AMPARs are expressed, like many other ion channels, invarious subunit combinations; they assemble from foursubunits, GluR1–4 (or GluRA–D), into ion channel tetra-mers [1,6–9]. The selectivity of the assembly process,which determines channel functional properties, is poorlyunderstood. The subunit stoichiometry also regulates

Corresponding author: Greger, I.H. (ig@mrc-lmb.cam.ac.uk).Available online 16 July 2007.

www.sciencedirect.com 0166-2236/$ – see front matter � 2007 Elsevier Ltd. All rights reserve

vesicular traffic [1,10,11] and synapse-specific targetingof channel tetramers [12–14], and thereby directly imp-acts synaptic efficacy and plasticity. Moreover, accumu-lating evidence suggests that the AMPAR composition isnot static, but can be altered in response to certain inputs[15–21].

In analogy to K+ channels [22], AMPARs are thought toassemble as dimers of dimers [9,23,24]. The N-terminaldomain mainly mediates dimer formation; subsequenttetramerisation involves the extracellular S2 loop andthe transmembrane segments (including the pore loop;Figure 1a) [23]. Receptor assembly occurs at the endo-plasmic reticulum (ER) membrane. Exit from the ER isunder stringent quality control, which monitors correctsubunit folding and assembly [25]. The efficiency ofthese processes impacts on ER export kinetics and ther-eby determines the number of channels available forexpression at synapses; this will ultimately tune theresponsiveness of a neuron. Interestingly, recent studiesindicate that conformational alterations associated withgating motions, such as ligand binding and desensitiza-tion (i.e. channel closure in ligand bound state), take placein the ER lumen, where they are sensed by the qualitycontrol machinery [26–31]. These findings add iGluRs to agrowing list of ER client proteins that require ligands or‘chemical chaperones’ for efficient folding and export fromthe ER [32].

In addition to the subunit stoichiometry, AMPARfunction is diversified further by RNA processing events,including alternative splicing and RNA editing [1]. Spli-cing of mutually exclusive exons (termed flip/flop) withinthe ligand-binding domain (LBD) modulates desensitiza-tion kinetics [33]. These two alternative exons are pre-sent in all four subunits and are common to vertebrateAMPARs (Figure 2) [34]. Also, RNA editing at two sitesmodulates AMPAR function — the R/G site within theLBD of GluR2–4 and the Q/R site in the pore loop ofGluR2 (Figure 1a; Box 1) [35,36]. These sites are locatedwithin subunit interfaces (Figure 1b) and intimatelyimpact receptor assembly [30,37]. Below, we discussthe role of individual domains in the assembly processand review structural parameters of their interactionsurfaces. We then describe how these interfaces aremodulated by RNA editing and how these modificationsevolved in the iGluR lineage (Figure 2; Box 2). Finally,we consider how gating motions, initiated by ligandbinding in the ER, contribute to receptor folding andassembly.

d. doi:10.1016/j.tins.2007.06.005

Figure 1. Domain organization of AMPARs and associated interfaces. (a) Schematic of the GluR2 subunit. The N-terminal LIVBP-like domain is in pink. The bipartite LBD

(yellow) is composed of the S1 (purple) and the S2 segments (green). The R/G editing site at position 743 is indicated by a red diamond, the flip/flop exon by a blue curve.

These two extracellular domains face the ER lumen (depicted by the grey vertical bar) during receptor biogenesis. The ion channel domain within the lipid bilayer (blue)

consists of three transmembrane (TM) segments (grey) and the re-entrant pore loop. The Q/R editing site at the pore apex is depicted by a red diamond. (Amino acid

positions are numbered according to the mature polypeptide lacking the signal sequence; therefore, in disagreement with [37], the Q/R site is labelled 586.) Two loops (25

and 10 residues in length) and the C terminus face the cytoplasm. (b) Structural depiction of an AMPAR complex. The background colour coding of the individual domains is

as in (a). A hypothetical subunit dimer is shown in cyan; the yellow dimer lacks the extracellular portion for clarity. The vertical arrow runs through the twofold symmetric

LIVBP and LBD axis, and the fourfold symmetric ion channel axis. Symmetry relations (i.e. twofold and fourfold) are indicated by symbols within the figure. The LIVBP-like

domain dimer is represented by the homologous extracellular portion of mGluR1 (PDB code 1EWK) [45]. The LBD dimer is shown in its R/G unedited L-Glu-bound form (PDB

code 2UXA) [30]. Arg743 sidechains are shown in stick representation and are located at a twofold symmetric subunit interface (red diamond). The vertical grey bar denotes

the ER lumen, as in (a). The fourfold symmetric ion channel domain is derived from a homology model [37] based on the atomic coordinates of the prokaryotic KcsA K+

channel. The long a helix corresponds to TM3, the short helix to the ascending part of the pore loop; TM1 and TM4 were not included in the model [37]. Q/R site residues

from each subunit are shown in stick form; they are located at a fourfold symmetric subunit interface (red diamond). The C-terminal domain is not included in the model. (c)

Back view of an LBD protomer. The S1 (purple) and S2 (green) segments comprise the bipartite LBD [30]. One LBD protomer is shown from the back, with a view onto the

elements forming the dimer interface (D, J, b). The Arg743–Asp490 salt bridge, which contributes to S1–S2 interactions, is shown. Cys773 at the C-terminal end is linked to

lower lobe Cys718 (indicated in orange); this cysteine bridge is required for folding [30]. The extreme N- and C-terminal ends of the LBD are indicated by black arrows, and

the (di-peptide) linker replacing TM1–TM3 [47] is depicted. Tertiary folding of this domain requires translation and ER translocation of the intervening core ion channel

(TM1–TM3); it is therefore expected to occur relatively late, probably after the N-termini have engaged in dimer formation.

408 Review TRENDS in Neurosciences Vol.30 No.8

Receptor assembly in the ER involves distinctdomainsThe biogenesis of oligomeric transmembrane proteinsoccurs at the ER membrane and commences with the co-translational insertion of nascent polypeptides throughthe Sec61 channel [38]. Subsequent protein folding andassembly is mediated by three physicochemically distinctenvironments: the cytosol, the lipid bilayer of the rough ERand the ER lumen. Folding of a nascent peptide into itstertiary structure can begin as early as the emergence ofthe polypeptide from the ribosome, as has been shown forthe Kv1.3 K+ channel [39]. Individual domains of eukar-yotic proteins appear to acquire their tertiary structuresequentially, which is thought to minimise misfolding ofmultidomain polypeptides [40].

iGluR subunits can be divided into four distinct domains(Figure 1a,b). The extracellular (and thus the ER-luminal)

www.sciencedirect.com

portion comprises two domains that are homologous toperiplasmic binding proteins (PBPs; [41]): the N-terminalLIVBP (leucine/isoleucine/valine-binding protein-likedomain) and the adjacent LBD, which is homologous toglutamine-binding protein (Figure 1b). The LBD domain isattached to the membrane-embedded ion channel andorchestrates gating (Figure 1b). The ion channel domainitself is composed of three transmembrane segments and are-entrant pore loop, which is distantly related to K+

channel pore loops [42]. The cytoplasmic C-terminaldomain interacts with various components of the postsyn-aptic signalling apparatus, which mediate receptor traf-ficking and anchorage [1,10,11]. During assembly, themonomeric subunits are likely to ‘zip up’ along the extra-cellular twofold symmetric surfaces in a vectorial fashion,that is, from the N to C terminus. The second assemblystep, tetramerisation, appears to be guided by contacts

Box 1. RNA editing.

RNA editing is an enzymatic modification that alters RNA composi-

tion by substrate-specific nucleotide insertion, deletion or alteration

[84]. Different types of editing have been described in eukaryotes.

Editing of mRNA can alter the sequence of the encoded protein,

whereas editing within non-coding regions, such as untranslated

regions (UTRs) or the introns of pre-mRNA, can affect RNA stability,

splicing and RNA interference [36]. In addition, RNA editing is

widespread and essential in the processing of tRNA, rRNA and

micro RNA (miRNA) [36,84,85].

The most prevalent modification in higher eukaryotes is the

deamination of adenosine (A) residues to inosine (I) [36]. A-to-I

editing was first detected as the unwinding of double-stranded RNA

(dsRNA). The enzymes responsible have since been cloned in

vertebrates: adenosine deaminases acting on RNA (ADARs) 1–3 [76].

Their significance is clearly demonstrated with mouse mutants:

mice lacking ADAR1 are embryonical lethal [86], whereas ADAR2

knockouts are prone to seizures and die within 3 weeks of birth [87].

In addition, loss-of-function mutants of orthologous genes in

Caenorhabditis elegans (Adr1 and Adr2) and Drosophila (dAdar)

show abnormal behaviour, and gain of dAdar function is lethal [36].

Many ADAR substrates are transcripts of genes expressed in the

nervous system [88], including vertebrate AMPAR [35]. Editing in

exon 11 replaces a glutamine (Q) with an arginine (R) in the pore

loop lining the channel (Figure 1a). Q/R editing reduces channel

conductance and inward rectification and lowers calcium perme-

ability. This editing normally occurs with almost 100% efficiency and

genomically encoding the arginine residue rescues the ADAR2 null

phenotype. A developmentally regulated editing site is located in

exon 13 of GluR2–4, where a glycine (G) replaces an arginine residue

within the LBD, resulting in quicker recovery of AMPARs from

desensitization [64]. Both editing sites also alter the biogenesis and

trafficking properties of the receptor [30,37]. Editing of AMPARs

requires an imperfect inverted repeat sequence in the intron

downstream of the editing sites. This editing complementary

sequence (ECS) folds back to give dsRNA, which is a substrate

requirement for ADARs [35,36]. The importance of editing at these

sites is also evident from the conservation of the ECS in vertebrate

evolution (Box 2; Figure 2) [89].

Box 2. Evolution of AMPAR RNA processing at subunit

interfaces.

RNA processing within the LBD of the functionally dominant GluR2

subunit is highly conserved in vertebrates (Figure 2). AMPAR-like

genes in protostomes and primitive chordates are devoid of these

processing events. The vertebrate AMPAR gene family is thought to

have evolved from a common ancestral subunit in the early

chordate lineage (Figure I). The urochordate Ciona intestinalis has

a single AMPAR-like subunit with only one exon as the primordial

vertebrate flip/flop exon. Hagfish, a primitive jawless vertebrate, has

two subunits, both of which have flip and flop exons, as do all

subunits of jawed vertebrates. An exon duplication event was

proposed to have occurred after the separation of vertebrates from

primitive chordates [34]. Functional diversification and alternative

splicing of the flip/flop exons probably evolved before the initial

expansion of the gene family.

In contrast to hagfish, jawed vertebrates have additional AMPAR-

like subunits. The sequence elements required for R/G editing in

GluR2–4 show striking conservation, indicating that R/G editing may

have appeared during early expansion of the gene family [89]. As

shown in Figure 2, the greatest divergence occurs in the terminal

pentaloop of the dsRNA substrate. This loop adopts a novel fold [90]

and mutations in this sequence affect the efficiency of editing [91].

No vertebrates have so far been discovered with the edited residue

genomically encoded.

Constitutive RNA editing occurs at the Q/R site in the GluR2

subunit of sharks and tetrapods (Figure 2) [92]. In some bony fish,

this intron sequence has diverged in GluR2 genes (e.g. in Tilapia, an

arginine is genomically encoded at the Q/R site). Interestingly, the

hagfish GluR2 subunit is intronless in the region encoding the pore

loop and has a genomically encoded arginine, indicating that Q/R

editing evolved in vertebrates after the separation from jawless fish

and that there is strong evolutionary selection for this residue in the

GluR2 subunit [92].

Figure I. Simplified taxonomic relations within the early chordate lineage.

Review TRENDS in Neurosciences Vol.30 No.8 409

within LBD dimers and generates the fourfold symmetricion channel (Figure 3a) [30]. Below, we consider currentinsights into the role of individual domains during subunitassembly, starting from the N-terminal domain.

The N-terminal LIVBP domainThe extracellular portion, which comprises �80% of asubunit, faces the ER lumen, where it folds and engagesin the initial steps of subunit assembly. The N-terminalLIVBP-like domain, which shares 23% sequence identitywith the glutamate-binding domain of the metabotropicreceptor mGluR1 (comparing rat GluR2 and rat mGluR1),is believed to determine interaction compatibility by ensur-ing association of subunit monomers only from a giveniGluR subfamily [23]. In the case of AMPARs, dimerisationof heteromers appears to be preferred over homodimerisa-tion [43]. As isolated LIVBP domains form stable dimers insolution [44], this initial interaction is likely to be tight.The crystal structure of the mGluR1 binding core dimerreveals a largely hydrophobic subunit interface, which isformed mainly by two helices (B and C) within the upperlobes of the clamshell-shaped domain [45]. Similarly,upper-lobe segments of the AMPAR N-terminus have beenimplicated in subtype-specific heteromerisation [24].Homology modelling indicated that the dimer interface,

www.sciencedirect.com

formed by upper lobe segments of iGluRN-termini, is morehydrophobic in AMPARs, but contains charged residues inkainate receptor N-termini. This difference might underliethe subfamily-specific interactions mediated by thisdomain [24].

The �400-residue continuous LIVBP-like segment(Figure 1a) probably adopts its tertiary a,b fold whilethe remaining polypeptide is still being translated andintegrated into the lipid bilayer. Newly synthesizedN-termini could therefore associate with other conco-mitantly translocating nascent chains or assemble withpreviously synthesized AMPAR monomers. How this firstassembly step is achieved is unclear (as it is indeed formany other heteromeric ion channels) and raises the prin-cipal question of how subunits preferentially associatewith other members of the same subfamily rather thansimply homo-oligomerise. As a given mRNA is translatedby polysomes, a high concentration of a particular subunit

Figure 2. Evolutionary conservation of regulatory elements required for RNA processing in GluR2. A schematic of the GluR2 protein (from segment 1 [S1] to the C terminus)

is aligned with the 30 end of the human GluR2 gene. Colour coding of the protein is as described in Figure 1a. The intron/exon structure is annotated with the editing sites

(red diamonds), their corresponding ECS (red), and alternative splicing of flop (o) and flip (i) exons. Below the gene structure, a plot of sequence conservation is shown for

the vertebrate GluR2 gene. Sequences were obtained from precomputed alignments of genomes provided by University of California Santa Cruz or Joint Genome Institute,

using the VISTA browser [93] and human as the base genome. Alternatively, sequences were found using BLAT ([94]; http://genome.ucsc.edu/]) or WU-BLAST 2.0 in

Ensembl (http://www.ensembl.org/index.html) of the current genome build using exons 11–16 (amino acids 492–883) of the human GluR2 peptide (P42262) or the gene

nucleotide sequence of the inverted repeat. Species marked with an asterisk had sequence for the whole gene region of interest and were used to create a conservation plot:

sequences were aligned using mLAGAN ([95]; http://lagan.stanford.edu/lagan_web/index.shtml) and the plot was visualized with SinicView [96]. Multiple sequence

410 Review TRENDS in Neurosciences Vol.30 No.8

www.sciencedirect.com

Figure 3. Model depicting the vectorial ‘zipping-up’ of interfaces during subunit assembly. (a) Steps encompassing dimer (I) and tetramer (II) formation are shown (see

text). Individual steps that could occur during dimerisation (I) are labelled a–d. Initially, LIVBP domains will interact (step Ib), which might be reversible (step a), depending

on the affinity of the interaction partners. In step Ic, the LBD dimer interface forms; stabilization of this interface may promote alignment and/or packing of ion channel

helices to form tetramerisation-competent dimers (step Id). This step is facilitated by Arg743 and L483Y. Alignment of dimers during tetramerisation (II) will involve the

packing of transmembrane helices. Re-entrant pore loops might flip into the lipid bilayer (schematised in b) to form the pore. Tetramer formation is likely to involve new

interfaces between subunits within the extracellular portion of the receptor; these interfaces and their overall symmetry are poorly understood at present (see [54,55]). ER-

export competence appears to require relaxation into a desensitized-like state (IIb) ([30]; see also [31]). This step is facilitated by N754D, a GluR2 mutation that destabilizes

the LBD interface and accelerates desensitization. Acquisition of this conformation may itself be an intermediate step, that is, it could be specifically recognized by a trans-

acting factor, such as Stg, that is required for forward traffic (IH Greger et al. unpublished observations). Colour coding is as in Figure 1b. (b) The potential flipping of pore

loops into the lipid bilayer may occur at the level of a dimer or during tetramerisation. As pore loops do not traverse the bilayer and therefore do not represent bona fide

transmembrane segments, they might equilibrate into lipid via a Sec61-independent mechanism. This process could be affected by the editing state of the Q/R site and by

the strength of LBD dimer contacts. Colour coding of the ion channel domain is as in Figure 1a.

Review TRENDS in Neurosciences Vol.30 No.8 411

is expected to accumulate at the ER membrane, in atemporally and spatially restricted fashion. In addition,diffusion is limited by the lipid bilayer and probably also byinteraction with ER-lumenal chaperone networks, whichwill restrict ‘collision’ of spatially separated translocationproducts even further. Together, these conditions are exp-ected to facilitate the formation of homo-oligomers. How-ever, initial dimerisation of homomeric subunits might beof lower affinity than dimer contacts between heteromersand be reversible once a heteromeric partner becomesavailable (Figure 3a; step Ia). Increasing the ER-dwelltime of certain subunits, and thus their availability, couldalso facilitate heteromerisation. This strategy appears tobe employed by the GluR2 subunit, which stably resides inthe ER (with a half-life of �12 h) [46]. This functionally

alignments (MSAs) of the editing sites were supplemented with additional sequences (li

using JalView. Additional sequences can be found under the following Genbank access

AF350049, AF350050, AF350051, AF350052; R/G site: AAFC03068020, AAGU01276109, A

tetrapod MSA for exon and intron 11 is represented as a consensus sequence. Em

respectively. Note that most sequence variation in the inverted repeat sequences and

stable G–U wobble pairs. Also note extensive intronic sequence conservation flanking

www.sciencedirect.com

critical subunit is present in the majority of AMPARtetramers in the brain [1].

The ligand-binding domainThe LBD lies C terminal to the LIVBP, and is split into S1and S2 segments by the ion channel. Tertiary folding of theLBD thus has to await translocation of the transmembranesegments (Figure 1a,c). Crystal structures of this domainare available for all three iGluR subfamilies [47–51], anddetails of agonist and antagonist interaction with thisdomain have been reviewed recently [51,52]. Ligand bind-ing to the crevice between the two lobes of the domain,known as D1 andD2, results in domain closure (Figure 4a).The resulting high-energy transition state is believed toproceed, in parallel, to either channel opening or desensi-

sted below), aligned using ClustalW (http://www.ebi.ac.uk/clustalw/) and visualized

ion numbers. Q/R site: AAFC03056666, AAGU01617172, AAGV01116826, AF350048,

AGV01116830, AF201347, AF201348; goldfish sequences can be found in [97]. The

pty triangles and arrows denote 50-splice sites and imperfect inverted repeats,

ECS is matched by complementary mutations or gives rise to thermodynamically

each of the flip/flop exons.

Figure 4. Potential impact of RNA editing on interface contacts. (a) Side view of an LBD dimer (flip splice form; [30]) harbouring Arg743 at the R/G site. Subunit protomers

are related by a twofold symmetry axis (dashed arrow) and are colour coded (cyan and yellow). D1 and D2 denote the upper and lower lobes of the LBD clamshell; the L-Glu

ligand (not shown in the figure) is engulfed between D1 and D2. Residues shown in magenta (Thr744, Pro745, Ser754 and Val758) are specific for the flip exon. The position

of the interface stabilizer L483Y [53] at the base of helix D is indicated by a red disc. Arg743, projecting from helix J at the top of the interface, is shown in stick form and is

circled. The right panel shows a close-up view of the Arg743 sidechains projecting across the dimer interface; terminal NH2 groups are separated by 3.4 A (0.34 nm). This

interface rearranges upon desensitization [54,56]. (b) Side view of subunit dimers at the level of the ion channel. The underlying homology model, which is based on KcsA

coordinates, encompasses TM3 (long helices) and Arg586 pore loops (short ascending helix plus descending coil; Arg586 sidechains are shown) [37]. The figure illustrates

that tetramer formation by two dimers harbouring Arg586 pore loops is of higher energetic cost (compared with Gln586 pore loops). This step may involve the flipping of

pore loops into the lipid bilayer (see Figure 3b), and would be faster for R–Q586 heterotetramers and Q586 homomers. (c) Top view onto the pore region, at the level of the Q/

R site. This homology model contains four glutamine residues at the Q/R site; the pore size was adjusted to match the sizes of permeant cations [37,67]. During pore

formation, Gln586 sidechains might engage mainchain carbonyls from the adjacent subunit in hydrogen bonding [67].

412 Review TRENDS in Neurosciences Vol.30 No.8

tization [53]. In all LBD structures, twofold symmetricaldimers were observed, with the two protomers arranged ina back-to-back fashion (Figure 4a). The dimer interface,which buries around 1150 A2 of solvent-accessible surfacearea per protomer in GluR2, is formed by the upper lobes,and involves helices D and J and the first interdomain b

strand (Figure 1c); interface contacts aremediated by polarand hydrophobic interactions [47].

In contrast to the LIVBP, isolated LBDs do not exist asdimers in solution (at concentrations below 1 mM) [54],revealing a much looser interaction in this part of the re-ceptor. The LBD constitutes the ‘muscle’ of the receptor andconveys gating motions to the channel in response to ligandbinding/unbinding, underlining the necessity of confor-mational flexibility in this region. Tightening of the inter-facebymutation orby theallostericmodulator cyclothiazidedramatically attenuates desensitization, suggesting thatdesensitization involves interface opening [52,54,55]. Speci-fically, mutation of Leu483 to tyrosine (in GluR2) createsfavourable cation–p interaction across the interface, whichstabilizes dimers dramatically [54] (Figure 4a). Using aGluR2 mutant that constrains the receptor in a desensi-tized-like state (S729C), Armstrong et al. recently examinedthe extent of LBD rearrangement upon desensitization [56].When comparedwith an ‘undesensitized’ (L483Y) structure,they observed complete rupture of the D1 interface andformation of a new interface involving the D2 segments.

www.sciencedirect.com

Asthe ionchannel is attached toD2, the increasedproximityof the lower lobes suggested a mechanism for channelclosure upon desensitization [56].

Low-resolution single-particle electron microscopyimages of native AMPARs demonstrated that exposure toglutamate resulted in a large proportion of receptors withwidely separatedN-termini, compared to a preparation thatalso included the desensitization blocker cyclothiazide.These data reflect the profound conformational changeinitiated by LBD dimers upon desensitization [57]. Recentexperiments with kainate receptors reveal that tighteningof D1 interface contacts via cysteine linkages similarlyattenuates desensitization for this iGluR subfamily, imply-ing that a related mechanism underlies desensitization[31,58]. However, the molecular determinants of this pro-cess differ between AMPA and kainate receptors [59,60]. Itis worth noting that AMPARs lacking the LIVBP domainproduce functional homomeric receptors when expressed inheterologous cells [61,62]. GluR0, a prokaryoticK+-selectiveiGluR, lacks the LIVBP domain. The GluR0LBD exists as adimer in solution, with a Kd of �800 nM [63]; the tightercontacts might compensate for the lack of the N-terminaldomain. In contrast to GluR2, the dimer interface is closedtowards its N-terminal end, where additional interdomainhydrogen bonds are formed (in the vicinity of theR/Geditingsite). GluR0 desensitization kinetics are about three ordersof magnitude slower than AMPAR kinetics [7]. However,

Review TRENDS in Neurosciences Vol.30 No.8 413

GluR0 is likely to exist as a homotetramer and thus mightnot require the ‘selector’ function exhibited by AMPARLIVBPs; neither does this channel require the fast kineticsthat are essential for synaptic AMPARs. The GluR0 LBDappears to combine both the assembly and gating functions.

The LBD is attached to the ion channel and orchestratesgating (Figure 1b). Assembly of the ion channel domain isinfluenced by LBD contacts and by RNA editing; thesesteps are discussed in the section below.

RNA editing, interface contacts and GluR2 assemblypropertiesAs mentioned above, the LBD interface in AMPARs ismodified by flip/flop splicing and R/G editing; both eventsmodulate gating kinetics [33,35,64,65] and receptor assem-bly [30,37]. Editing at the R/G site of GluR2–4 results in adramatic change in the sidechain chemistry at the top ofthe interface, where adenosine deaminase acting on RNA(ADAR) action replaces Arg743 with glycine (Figures 2,and 4a). This modification, which endows the receptor withfunctional diversity, is unique to vertebrate AMPARs andis accomplished by a highly conserved intronic sequenceimmediately downstream of the editing site (Box 1;Figure 2). Most other vertebrate and invertebrate iGluRsubtypes constitutively express arginine at this position;an exception is squid, which harbours a genomicallyencoded glycine in the two non-NMDA-like subunits clonedso far [66]. How does the R/G modification affect interfacecontacts? The X-ray structure of the unedited LBD (in a flipbackground) revealed an unexpected arrangement of theArg743 rotamers across the dimer interface. The side-chains were observed in an extended conformation, pro-jecting towards one another to within 3.4 A (rather thanpointing upward toward the solvent; Figure 4a) [30]. Thedistance between their Ca atoms across the interface is�1 A less than that seen in the edited glycine structure(solved in the flop background [47]). Together, these struc-tural data indicated an unconventional Arg–Arg inter-action facilitated by surrounding negative charges, suchas the highly conserved Asp490 sidechain (Figure 1c) [30].Whether this interaction actually stabilizes dimers is cur-rently unclear. Functional data do not fully resolve thisissue, as the kinetics of entry into desensitization dependon a given subunit and on the flip/flop splice form. They areindeed slower for flop Arg743 isoforms (as would beexpected for an interface stabilizer) but are faster for fliparginine isoforms [65]. In kainate receptors, the equivalentarginine (Arg775) forms a salt bridge with an adjacentaspartate (Asp776) of the opposite subunit [49,50]. Thiselectrostatic glue does not appear to stabilize the interface,as judged by the high Kd value obtained from analyticalultracentrifugation runs (>6 mM) [58] and by the fastdesensitization kinetics characteristic of kainate receptors[58,59].

Interestingly, the GluR2 Arg743 form has a greatertendency to (homo)tetramerise; it also folds more effi-ciently and buds from the ER at higher rates than theGly743 form [30]. These effects were observed in fourGluR2 backgrounds (i.e. flip/flop and Q/R), both in neuronsand in heterologous cells. Previous experiments had showna role for RNA editing in AMPAR assembly, whereby

www.sciencedirect.com

editing at the Q/R site attenuated tetramerisation of fullyedited GluR2 flop homomers in neurons and restrictedsubsequent ER export [37]. The Q/R site lies at a four-fold symmetric interface, at the apex of the pore loop(Figure 1b); it was suggested that juxtaposing four argi-nine-containing pore loops during the second assemblystep, the dimerisation of dimers, was rate limiting (Figures3b and 4b). Testing a panel of amino acids at this positionrevealed that most efficient tetramerisation was achievedwith glutamine and tryptophan at the pore apex (hydro-phobic residues were not tested) [37]. The fact that assem-bly proceeded more efficiently with glutamine than withasparagine indicated that, apart from charge, a specificsidechain length is critical. Four glutamine sidechains, onefrom each pore loop, could engage in favourable hydrogenbonding (Figure 4c) [67], whereas tryptophan residuesmight engage in stabilizing stacking interactions duringpore formation [37].

As mentioned, the energy barrier of arginine poreformation is lowered by alterations of LBD dimer contacts,as Q/R-edited GluR2 containing Arg743 has a greatertendency to tetramerise than its Gly743 counterpart[30]. In addition, the interface stabilizer L483Y promotestetramerisation of fully edited GluR2, perhaps by aligningsubunits within a dimer (Figure 3a; step Id). WhetherArg743 contacts indeed stabilize intersubunit contacts isat present unclear. Arg743 could also facilitate LBD fold-ing by linking the S1 and S2 segments; Arg743 forms a saltbridgewith Asp490, which itself projects from a loopwithinS1 (Figure 1c). Together, these data revealed that RNAediting alters GluR2 assembly properties: unedited GluR2is competent to form homomers, whereas editing restrictsself-assembly, implying that fully edited GluR2 preferen-tially heterotetramerises (Figure 5).

ER exit of tetramers and functional quality controlL483Y GluR2 mutants tetramerise, but exit from the ERinefficiently [30]. Stabilizing the LBD interface thus over-comes a rate-limiting step in the assembly pathway, butthis step needs to be reversible, that is, once assembled thechannel might have to collapse into a low-energy, desensi-tized-like state (Figure 3a; step IIb). This conformationaltransition appears to render the channel ER-export com-petent [30]; it could result in chaperone dissociation orfacilitate association with a transport factor. A potentialcandidate for the latter scenario is the AMPAR-auxiliarysubunit stargazin (Stg; Figure 3a) [68]. Stg and itsrelatives, the transmembrane AMPA receptor regulatoryproteins (TARPs), are required for early secretory traffick-ing [69]; they interact with the LBD, slow gating kineticsand appear to recognize specific conformational states ofthe LBD [68,70]. N754D, a GluR2 mutant that acceleratesdesensitization, promoted budding from the ER in neurons[30], but not in Hek293 cells (authors’ unpublished obser-vations). Thismight reflect that Stg, which is not expressedin Hek293 cells, mediated efficient trafficking of N754DGluR2 in neurons. In addition, we noticed that GluR2harbouring the lurcher mutation A622T, which reducesdesensitization of GluR1 [71], produced a retention phe-notype (authors’ unpublished observations). Recent exper-iments revealed that kainate receptors, when locked in the

Figure 5. Impact of RNA editing on the assembly behaviour of GluR2. In the

current model, unedited GluR2 (Q/R, R/G; step iv) is competent to oligomerise into

GluR2 homotetramers [37]. GluR2 (Q) assembles regardless of the editing state at

the R/G site, although unedited GluR2 traffics more efficiently than partially edited

GluR2 (i.e. Q/R, R/G GluR2) [30]. Editing at the Q/R site severely restricts

homotetramer formation of doubly edited GluR2 (Q/R, R/G; step ii), which could

encourage heteromer formation (step iii), that is, fully edited GluR2 appears to

require co-assembly with GluR1, GluR3 or GluR4 for efficient channel formation

and subsequent ER exit. Singly edited GluR2 (at the Q/R site only: Q/R, R/G; step

i) retains limited assembly competence. It should be noted that fully edited

GluR2 will tetramerise, as these channels produce current when expressed

heterologously; however, assembly is expected to be slow and have a high

energetic cost. In addition, resulting R586 tetramers are less stable.

414 Review TRENDS in Neurosciences Vol.30 No.8

open state, are also retained in the ER [31]; as kainatereceptors do not interact with Stg, other components of theERquality controlmachinerywill be involved insensingER-export-competent, quaternary conformations of iGluRs.

These data, together with the finding that ligandbinding appears to be required for ER transit [29], indicatethat conformational changes underlying gating motions(i.e. resting-open-desensitized) are sensed during iGluRbiogenesis in the ER. L-Glu, which is highly concentratedin the cytosol [29], could reach luminal AMPARs throughthe relatively ‘leaky’ ER membrane [72] or via Glu trans-porters. The latter possibility has been reported for intra-cellular mGluR5, which signals at endo-membranes inresponse to ligand binding [73]. Inability to bind ligandalters receptor folding properties [30]. At what stage ligandbinding feeds into the folding and assembly pathway iscurrently unclear; as a prerequisite for desensitization, it is

www.sciencedirect.com

expected to act before this conformational alteration. L-Glucould also bind to assembly intermediates, such as subunitdimers, and influence downstream assembly steps. In con-trast to the cysteine-loop ion channel family, individualiGluR subunits can bind ligand individually [6], which givesrise to various subconductance states and complex kineticgating schemes (e.g. [53]). Altering the rate constants be-tween any of these states will shift thermodynamic equili-bria between conformers, which could influence channelformation and, in turn, ER exit kinetics (Figure 3a). RNAediting sites and the flip/flop splice site ([74]; I.H. Greger,unpublished), which line subunit interfaces and alter gatingkinetics, are ideally suited to influence AMPAR biogenesis.In fact, the finding that flip splice forms traffic more effi-ciently in Hek293 cells [74] might partly explain why flipchannels display functional dominance over flop forms inthis system [75].

OutlookRNA processing within the LBD is developmentallyregulated [35,64]. As these changes also alter assemblyproperties [30], they could encourage formation of GluR2-containing tetramers during development (Figure 5).ADARs, the enzymes responsible for adenosine-to-inosineediting (Box 1; [76]), are under sophisticated negative feed-backcontrol [36,77,78]andare regulatedbydiverseexternalcues [79–81]. If alterations in neuronal network activitywere to alter ADAR levels, regulation of AMPAR assemblycould be envisaged. As editing is manifested at the level ofsubunit mRNA, the resulting change in post-synapticAMPAR responsiveness would most likely follow a slowtime course and might play a role during long-term adjust-ments, suchas thoseoccurringduringhomeostaticplasticity[3].

AMPAR assembly does not seem to operatestochastically. Different tetramer combinations can betargeted to selected dendritic subdomains within a givenneuron [12–14]. In addition, synapses can ‘employ’ func-tionally different tetramers in response to altered inputpatterns [15–21,82]; this regulation might involve localsynapse-specific remodelling of AMPARs [83]. Taken toge-ther, these findings demonstrate that neurons are able totailor AMPAR complexes depending on their specificneeds, which will increase their computational capacitysubstantially. Whether these different tetramers are asse-mbled in response to altered activity patterns or, alterna-tively, whether ‘pre-assembled’ complexes are specificallytrafficked remains an open question.

AcknowledgementsWe are grateful to colleagues within the MRC LMB, as well as JonHanley, Marie Ohman and Mark Mayer, for comments on themanuscript. We thank Jo Westmoreland and Graham Lingley for helpwith the illustrations. We acknowledge the support of the Royal Society(IHG), the MRC (IHG and ACP) and NIH grant AG 13620 (EBZ).

References1 Dingledine, R. et al. (1999) The glutamate receptor ion channels.

Pharmacol. Rev. 51, 7–612 Malenka, R.C. and Bear, M.F. (2004) LTP and LTD: an embarrassment

of riches. Neuron 44, 5–213 Turrigiano, G.G. and Nelson, S.B. (2004) Homeostatic plasticity in the

developing nervous system. Nat. Rev. Neurosci. 5, 97–107

Review TRENDS in Neurosciences Vol.30 No.8 415

4 Jonas, P. (2000) The time course of signaling at central glutamatergicsynapses. News Physiol. Sci. 15, 83–89

5 Wanamaker, C.P. et al. (2003) Regulation of nicotinic acetylcholinereceptor assembly. Ann. N. Y. Acad. Sci. 998, 66–80

6 Rosenmund, C. et al. (1998) The tetrameric structure of a glutamatereceptor channel. Science 280, 1596–1599

7 Chen, G.Q. et al. (1999) Functional characterization of a potassium-selective prokaryotic glutamate receptor. Nature 402, 817–821

8 Safferling, M. et al. (2001) First images of a glutamate receptorion channel: oligomeric state and molecular dimensions of GluRBhomomers. Biochemistry 40, 13948–13953

9 Tichelaar, W. et al. (2004) The three-dimensional structure of anionotropic glutamate receptor reveals a dimer-of-dimers assembly. J.Mol. Biol. 344, 435–442

10 Barry, M.F. and Ziff, E.B. (2002) Receptor trafficking and the plasticityof excitatory synapses. Curr. Opin. Neurobiol. 12, 279–286

11 Bredt, D.S. and Nicoll, R.A. (2003) AMPA receptor trafficking atexcitatory synapses. Neuron 40, 361–379

12 Toth, K. and McBain, C.J. (1998) Afferent-specific innervation of twodistinct AMPA receptor subtypes on single hippocampal interneurons.Nat. Neurosci. 1, 572–578

13 Rubio, M.E. and Wenthold, R.J. (1997) Glutamate receptors areselectively targeted topostsynaptic sites inneurons.Neuron18, 939–950

14 Gardner, S.M. et al. (2001) Correlation of AMPA receptor subunitcomposition with synaptic input in the mammalian cochlear nuclei.J. Neurosci. 21, 7428–7437

15 Liu, S.Q. and Cull-Candy, S.G. (2000) Synaptic activity at calcium-permeable AMPA receptors induces a switch in receptor subtype.Nature 405, 454–458

16 Liu, S.J. and Cull-Candy, S.G. (2002) Activity-dependent change inAMPA receptor properties in cerebellar stellate cells. J. Neurosci. 22,3881–3889

17 Cull-Candy, S. et al. (2006) Regulation of Ca2+-permeable AMPAreceptors: synaptic plasticity and beyond. Curr. Opin. Neurobiol. 16,288–297

18 Bagal, A.A. et al. (2005) Long-term potentiation of exogenousglutamate responses at single dendritic spines. Proc. Natl. Acad.Sci. U. S. A. 102, 14434–14439

19 Plant, K. et al. (2006) Transient incorporation of native GluR2-lackingAMPA receptors during hippocampal long-term potentiation. Nat.Neurosci. 9, 602–604

20 Goel, A. et al. (2006) Cross-modal regulation of synaptic AMPAreceptors in primary sensory cortices by visual experience. Nat.Neurosci. 9, 1001–1003

21 Bellone, C. and Luscher, C. (2006) Cocaine triggered AMPA receptorredistribution is reversed in vivo by mGluR-dependent long-termdepression. Nat. Neurosci. 9, 636–641

22 Tu, L. and Deutsch, C. (1999) Evidence for dimerization of dimers in K+

channel assembly. Biophys. J. 76, 2004–201723 Ayalon, G. and Stern-Bach, Y. (2001) Functional assembly of AMPA

and kainate receptors is mediated by several discrete protein-proteininteractions. Neuron 31, 103–113

24 Ayalon, G. et al. (2005) Two regions in the N-terminal domain ofionotropic glutamate receptor 3 form the subunit oligomerizationinterfaces that control subtype-specific receptor assembly. J. Biol.Chem. 280, 15053–15060

25 Ellgaard, L. and Helenius, A. (2003) Quality control in the endoplasmicreticulum. Nat. Rev. Mol. Cell Biol. 4, 181–191

26 Grunwald, M.E. and Kaplan, J.M. (2003) Mutations in the ligand-binding and pore domains control exit of glutamate receptors from theendoplasmic reticulum in C. elegans. Neuropharmacology 45, 768–776

27 Valluru, L. et al. (2005) Ligand binding is a critical requirement forplasmamembrane expression of heteromeric kainate receptors. J. Biol.Chem. 280, 6085–6093

28 Mah, S.J. et al. (2005) Glutamate receptor trafficking: endoplasmicreticulum quality control involves ligand binding and receptorfunction. J. Neurosci. 25, 2215–2225

29 Fleck, M.W. (2006) Glutamate receptors and endoplasmic reticulumquality control: looking beneath the surface. Neuroscientist 12, 232–244

30 Greger, I.H. et al. (2006) Developmentally regulated, combinatorialRNA processing modulates AMPA receptor biogenesis. Neuron 51,85–97

www.sciencedirect.com

31 Priel, A. et al. (2006) Block of kainate receptor desensitization uncoversa key trafficking checkpoint. Neuron 52, 1037–1046

32 Bernier, V. et al. (2004) Pharmacological chaperones: potentialtreatment for conformational diseases. Trends Endocrinol. Metab.15, 222–228

33 Sommer, B. et al. (1990) Flip and flop: a cell-specific functional switch inglutamate-operated channels of the CNS. Science 249, 1580–1585

34 Chen, Y.C. et al. (2006) The mutually exclusive flip and flop exons ofAMPA receptor genes were derived from an intragenic duplication inthe vertebrate lineage. J. Mol. Evol. 62, 121–131

35 Seeburg, P.H. et al. (1998) RNA editing of brain glutamate receptorchannels: mechanism and physiology. Brain Res. Brain Res. Rev. 26,217–229

36 Valente, L. and Nishikura, K. (2005) ADAR gene family and A-to-IRNA editing: diverse roles in posttranscriptional gene regulation.Prog. Nucleic Acid Res. Mol. Biol. 79, 299–338

37 Greger, I.H. et al. (2003) AMPA receptor tetramerization is mediatedby Q/R editing. Neuron 40, 763–774

38 Clemons, W.M., Jr et al. (2004) Structural insight into the proteintranslocation channel. Curr. Opin. Struct. Biol. 14, 390–396

39 Kosolapov, A. et al. (2004) Structure acquisition of the T1 domain ofKv1.3 during biogenesis. Neuron 44, 295–307

40 Netzer, W.J. and Hartl, F.U. (1997) Recombination of protein domainsfacilitated by co-translational folding in eukaryotes. Nature 388,343–349

41 Quiocho, F.A. and Ledvina, P.S. (1996) Atomic structure and specificityof bacterial periplasmic receptors for active transport and chemotaxis:variation of common themes. Mol. Microbiol. 20, 17–25

42 Wood, M.W. et al. (1995) Structural conservation of ion conductionpathways in K channels and glutamate receptors. Proc. Natl. Acad. Sci.U. S. A. 92, 4882–4886

43 Mansour, M. et al. (2001) Heteromeric AMPA receptors assemble witha preferred subunit stoichiometry and spatial arrangement. Neuron32, 841–853

44 Kuusinen, A. et al. (1999) Oligomerization and ligand-bindingproperties of the ectodomain of the alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor subunit GluRD. J. Biol.Chem. 274, 28937–28943

45 Kunishima,N. et al. (2000) Structural basis of glutamate recognition bya dimeric metabotropic glutamate receptor. Nature 407, 971–977

46 Greger, I.H. et al. (2002) RNA editing at Arg607 controls AMPAreceptor exit from the endoplasmic reticulum. Neuron 34, 759–772

47 Armstrong, N. and Gouaux, E. (2000) Mechanisms for activation andantagonism of an AMPA-sensitive glutamate receptor: crystalstructures of the GluR2 ligand binding core. Neuron 28, 165–181

48 Furukawa, H. and Gouaux, E. (2003) Mechanisms of activation,inhibition and specificity: crystal structures of the NMDA receptorNR1 ligand-binding core. EMBO J. 22, 2873–2885

49 Mayer, M.L. (2005) Crystal structures of the GluR5 and GluR6 ligandbinding cores: molecular mechanisms underlying kainate receptorselectivity. Neuron 45, 539–552

50 Nanao, M.H. et al. (2005) Structure of the kainate receptor subunitGluR6 agonist-binding domain complexedwith domoic acid.Proc. Natl.Acad. Sci. U. S. A. 102, 1708–1713

51 Mayer, M.L. (2006) Glutamate receptors at atomic resolution. Nature440, 456–462

52 Mayer, M.L. and Armstrong, N. (2004) Structure and function ofglutamate receptor ion channels. Annu. Rev. Physiol. 66, 161–181

53 Raman, I.M. and Trussell, L.O. (1995) The mechanism of alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionate receptor desensitizationafter removal of glutamate. Biophys. J. 68, 137–146

54 Sun, Y. et al. (2002) Mechanism of glutamate receptor desensitization.Nature 417, 245–253

55 Horning, M.S. and Mayer, M.L. (2004) Regulation of AMPA receptorgating by ligand binding core dimers. Neuron 41, 379–388

56 Armstrong, N. et al. (2006) Measurement of conformational changesaccompanying desensitization in an ionotropic glutamate receptor.Cell127, 85–97

57 Nakagawa, T. et al. (2005) Structure and different conformationalstates of native AMPA receptor complexes. Nature 433, 545–549

58 Weston, M.C. et al. (2006) Conformational restriction blocksglutamate receptor desensitization. Nat. Struct. Mol. Biol. 13, 1120–1127

416 Review TRENDS in Neurosciences Vol.30 No.8

59 Fleck,M.W. (2003) Amino-acid residues involved in glutamate receptor6 kainate receptor gating and desensitization. J. Neurosci. 23, 1219–1227

60 Zhang, Y. et al. (2006) Interface interactionsmodulating desensitizationof the kainate-selective ionotropic glutamate receptor subunit GluR6. J.Neurosci. 26, 10033–10042

61 Pasternack, A. et al. (2002) Alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor channels lacking the N-terminal domain. J. Biol. Chem. 277, 49662–49667

62 Matsuda, S. et al. (2005) Roles of theN-terminal domain on the functionand quaternary structure of the ionotropic glutamate receptor. J. Biol.Chem. 280, 20021–20029

63 Mayer, M.L. et al. (2001) Mechanisms for ligand binding to GluR0 ionchannels: crystal structures of the glutamate and serine complexes anda closed apo state. J. Mol. Biol. 311, 815–836

64 Lomeli, H. et al. (1995) Control of kinetic properties of AMPA receptorchannels by nuclear RNA editing. Science 266, 1709–1713

65 Grosskreutz, J. et al. (2003) Kinetic properties of human AMPA-typeglutamate receptors expressed in HEK293 cells. Eur. J. Neurosci. 17,1173–1178

66 Battaglia, A.A. (2003) Cloning and characterization of an ionotropicglutamate receptor subunit expressed in the squid nervous system.Eur. J. Neurosci. 17, 2256–2266

67 Tikhonov, D.B. et al. (2002) Modeling of the pore domain of the GLUR1channel: homology with K+ channel and binding of channel blockers.Biophys. J. 82, 1884–1893

68 Ziff, E.B. (2007) TARPs and the AMPA receptor trafficking paradox.Neuron 53, 627–633

69 Tomita, S. et al. (2003) Functional studies and distribution define afamily of transmembrane AMPA receptor regulatory proteins. J. CellBiol. 161, 805–816

70 Tomita, S. et al. (2004) Dynamic interaction of stargazin-like TARPswith cycling AMPA receptors at synapses. Science 303, 1508–1511

71 Klein, R.M. and Howe, J.R. (2004) Effects of the lurcher mutation onGluR1 desensitization and activation kinetics. J. Neurosci. 24, 4941–4951

72 Le Gall, S. et al. (2004) The endoplasmic reticulum membrane ispermeable to small molecules. Mol. Biol. Cell 15, 447–455

73 Jong, Y.J. et al. (2005) Functional metabotropic glutamate receptors onnuclei from brain and primary cultured striatal neurons. Role oftransporters in delivering ligand. J. Biol. Chem. 280, 30469–30480

74 Coleman, S.K. et al. (2006) Isoform-specific early trafficking of AMPAreceptor flip and flop variants. J. Neurosci. 26, 11220–11229

75 Brorson, J.R. et al. (2004) Selective expression of heteromeric AMPAreceptors driven by flip-flop differences. J. Neurosci. 24, 3461–3470

76 Keegan, L.P. et al. (2004) Adenosine deaminases acting on RNA(ADARs): RNA-editing enzymes. Genome Biol. 5, 209

77 Rueter, S.M. et al. (1999) Regulation of alternative splicing by RNAediting. Nature 399, 75–80

Free journals for dev

The WHO and six medical journal publishers hav

Research Initiative, which enables nearly 70 of the

to biomedical literature

The science publishers, Blackwell, Elsevier, Harc

International Health and Science, Springer-Verlag a

and the British Medical Journal in 2001. Initially, m

free or at significantly reduced prices to universit

institutions in developing countries. In 2002, 22 ad

journals are now available. Currently more than 7

Gro Harlem Brundtland, the former director-gen

‘‘perhaps the biggest step ever taken towards red

and poor co

For more information, vis

www.sciencedirect.com

78 Feng, Y. et al. (2006) Altered RNA editing in mice lacking ADAR2autoregulation. Mol. Cell. Biol. 26, 480–488

79 Yang, J.H. et al. (2003) Widespread inosine-containing mRNA inlymphocytes regulated by ADAR1 in response to inflammation.Immunology 109, 15–23

80 Gan, Z. et al. (2006) RNA editing by ADAR2 is metabolicallyregulated in pancreatic islets and beta-cells. J. Biol. Chem. 281,33386–33394

81 Peng, P.L. et al. (2006) ADAR2-dependent RNA editing of AMPAreceptor subunit GluR2 determines vulnerability of neurons inforebrain ischemia. Neuron 49, 719–733

82 Liu, S.J. and Zukin, R.S. (2007) Ca2+-permeable AMPA receptors insynaptic plasticity and neuronal death. Trends Neurosci. 30, 126–134

83 Sutton, M.A. et al. (2006) Miniature neurotransmission stabilizessynaptic function via tonic suppression of local dendritic proteinsynthesis. Cell 125, 785–799

84 Gott, J.M. and Emeson, R.B. (2000) Functions andmechanisms of RNAediting. Annu. Rev. Genet. 2000, 499–531

85 Nishikura, K. (2006) Editor meets silencer: crosstalk between RNAediting and RNA interference. Nat. Rev. Mol. Cell Biol. 7, 919–931

86 Wang, Q. et al. (2000) Requirement of the RNA editing deaminaseADAR1 gene for embryonic erythropoiesis. Science 290, 1765–1768

87 Higuchi, M. et al. (2000) Point mutation in an AMPA receptor generescues lethality in mice deficient in the RNA-editing enzyme ADAR2.Nature 406, 78–81

88 Hoopengardner, B. et al. (2003) Nervous system targets of RNA editingidentified by comparative genomics. Science 301, 832–836

89 Aruscavage, P.J. and Bass, B.L. (2000) A phylogenetic analysis revealsan unusual sequence conservation within introns involved in RNAediting. RNA 6, 257–269

90 Stefl, R. and Allain, F.H. (2005) A novel RNA pentaloop fold involved intargeting ADAR2. RNA 11, 592–597

91 Stefl, R. et al. (2006) Structure and specific RNA binding of ADAR2double-stranded RNA binding motifs. Structure 14, 345–355

92 Kung, S.S. et al. (2001) Q/R RNA editing of the AMPA receptor subunit2 (GRIA2) transcript evolves no later than the appearance ofcartilaginous fishes. FEBS Lett. 509, 277–281

93 Frazer, K.A. et al. (2004) VISTA: computational tools for comparativegenomics. Nucleic Acids Res. 32, W273–W279

94 Kent, W.J. (2002) BLAT - the BLAST-like alignment tool. Genome Res.12, 656–664

95 Brudno, M. et al. (2003) LAGAN and Multi-LAGAN: efficient tools forlarge-scale multiple alignment of genomic DNA. Genome Res. 13, 721–731

96 Shih, A.C. et al. (2006) SinicView: a visualization environment forcomparisons of multiple nucleotide sequence alignment tools. BMCBioinformatics 7, 103

97 Li, Z. et al. (1999) Goldfish brain GluR2: multiple forms, RNA editing,and alternative splicing. Brain Res. Mol. Brain Res. 67, 211–220

eloping countries

e launched the Health InterNetwork Access to

world’s poorest countries to gain free access

through the internet.

ourt Worldwide STM group, Wolters Kluwer

nd John Wiley, were approached by the WHO

ore than 1500 journals were made available for

ies, medical schools, and research and public

ditional publishers joined, and more than 2000

0 publishers are participating in the program.

eral of the WHO, said that this initiative was

ucing the health information gap between rich

untries’’.

it www.who.int/hinari