Ca2+‐permeable AMPA receptors and their auxiliary subunits in synaptic plasticity and disease

Abstract AMPA receptors are tetrameric glutamate‐gated ion channels that mediate a majority of fast excitatory neurotransmission in the brain. They exist as calcium‐impermeable (CI‐) and calcium‐permeable (CP‐) subtypes, the latter of which lacks the GluA2 subunit. CP‐AMPARs display an array of distinctive biophysical and pharmacological properties that allow them to be functionally identified. This has revealed that they play crucial roles in diverse forms of central synaptic plasticity. Here we summarise the functional hallmarks of CP‐AMPARs and describe how these are modified by the presence of auxiliary subunits that have emerged as pivotal regulators of AMPARs. A lasting change in the prevalence of GluA2‐containing AMPARs, and hence in the fraction of CP‐AMPARs, is a feature in many maladaptive forms of synaptic plasticity and neurological disorders. These include modifications of glutamatergic transmission induced by inflammatory pain, fear conditioning, cocaine exposure, and anoxia‐induced damage in neurons and glia. Furthermore, defective RNA editing of GluA2 can cause altered expression of CP‐AMPARs and is implicated in motor neuron damage (amyotrophic lateral sclerosis) and the proliferation of cells in malignant gliomas. A number of the players involved in CP‐AMPAR regulation have been identified, providing useful insight into interventions that may prevent the aberrant CP‐AMPAR expression. Furthermore, recent molecular and pharmacological developments, particularly the discovery of TARP subtype‐selective drugs, offer the exciting potential to modify some of the harmful effects of increased CP‐AMPAR prevalence in a brain region‐specific manner.


Introduction
The basic properties of AMPA-type glutamate receptors (AMPARs) shape many of the key features of fast excitatory transmission in the CNS. Together with NMDA-type glutamate receptors these ligand-gated non-selective cation channels are involved both in synaptic signalling and the induction of various forms of synaptic plasticity (Traynelis et al. 2010;Huganir & Nicoll, 2013;Greger et al. 2017). At many synapses, AMPAR changes are primarily responsible for the expression of plasticity. Most notably, changes in their number or function underlie the activity-dependent strengthening or weakening of synaptic contacts, as seen in the processes of long-term potentiation and depression, the homeostatic adjustments that maintain neuronal excitability, and many other important forms of plasticity (Diering & Huganir, 2018).
Here we give a brief overview of those forms of plasticity that involve a change in the synaptic expression of one particular class of AMPAR -the calcium-permeable AMPA receptors (CP-AMPARs). These have emerged as important participants not only in a variety of conventional plasticities, but also in detrimental forms that are implicated in various neurological conditions. Most, if not all AMPARs, are associated with transmembrane auxiliary proteins that influence the receptors' biogenesis, their post-and presynaptic localization at synapses, and their functional properties (Jackson & Nicoll, 2011b;Rigby et al. 2015;Greger et al. 2017;Schwenk et al. 2019). We have, therefore, focused our review on the results from studies aimed at identifying specific roles for transmembrane auxiliary proteins in normal and detrimental forms of CP-AMPAR regulation. As much of the work described here has depended on the identification of CP-AMPARs from their hallmark properties, we start by summarizing these, and the way in which they are modified when receptors are assembled with particular auxiliary proteins.

Functional hallmarks of pore-forming and auxiliary proteins
The properties of AMPARs, notably their kinetics and Ca 2+ permeability, reflect the nature of the receptors' constituent subunits (Traynelis et al. 2010) and auxiliary subunits (Jackson & Nicoll, 2011b). The main players are depicted in Fig. 1.
Subunit composition can vary across brain regions and between cell types and can change during development and in response to neuronal activity. Of the four homologous pore-forming subunits (GluA1-GluA4) the GluA2 subunit plays a particularly critical role in determining AMPAR behaviour. GluA2 pre-mRNA is subject to nucleotide editing (mRNA editing) that results in the conversion of a genetically encoded glutamine (Q) to an arginine (R) at position 607 -the Q/R site in the pore-forming loop of M2. This switch, from a neutral to a positively charged residue in the channel's ionic selectivity filter, means that unlike GluA2-lacking AMPARs those containing GluA2 are Ca 2+ -impermeable (Burnashev et al. 1992). Q/R editing within GluA2's pore loop is highly efficient and serves not only to control Ca 2+ permeability  subunits within the ATD, LBD and TMD layers when viewed from the top (extracellular surface) of the receptor, along the overall twofold axis of symmetry. Of note, the arrangement of core subunits in AMPARs is not as strict as seen in NMDARs (Greger et al. 2017), and for this native AMPAR the positioning of the GluA2 subunits differs from the A/C positions reported for the first recombinant heteromeric GluA2/3 structure (Herguedas et al. 2016). Nevertheless, the fourfold symmetry of the TMD layer is common to both. C, cartoon representation of the TMD layer arrangement for a Ca 2+ -impermeable (CI-) AMPAR containing Q/R edited GluA2 subunits and a GluA2-lacking Ca 2+ -permeable (CP-) AMPAR. D, schematic illustrations of AMPAR key auxiliary subunits. TARPs and GSG1L belong to the claudin superfamily and have four transmembrane α-helices (numbered) and similar overall structures. Type Ia (γ 2, 3) and Type 1b (γ 4, 8) TARPs have canonical TTPV PDZ binding motifs whereas Type II TARPs (γ 5, 7) have atypical PDZ binding motifs (SSPC and TSPC). Note that because the transmembrane helices form a bundle within the membrane the TM2/TM3 linker (dotted) is shorter than shown. CNIHs also have four transmembrane α-helices but both the N and C termini are extracellular (Nakagawa, 2019). CKAMPs have a single transmembrane α-helix, an extracellular cysteine-rich region (the cysteine knot) and a PDZ binding motif (EVTV). but also to increase the proportion of GluA2-containing surface receptors by limiting the exit of GluA2 from the endoplasmic reticulum except when associated with unedited subunits (Greger et al. 2002;2003).
In addition to the pore-forming subunits, proteins belonging to several distinct families have emerged as important AMPAR constituents (Fig. 1), acting as auxiliary subunits that influence the receptors' biogenesis and localization within the cell membrane, as well as their biophysical and pharmacological properties. Those that contribute to the proteomic 'core' of the receptor (Schwenk et al. 2012) include the transmembrane AMPAR regulatory proteins (TARPs; γ 2, -3, -4, -5, -7 and -8) (Jackson & Nicoll, 2011b), two widely occurring members of the cornichon family (CNIH2 and -3) (Schwenk et al. 2009;Nakagawa, 2019), and the germ cell-specific gene 1-like protein (GSG1L) (Schwenk et al. 2014;Shanks et al. 2014). Other protein families that contribute to the 'peripheral' components of the proteome include the   Washburn & Dingledine, 1996;Jackson et al. 2011Tsubokawa et al. 1995Magazanik et al. 1997 cysteine-knot AMPAR modulating proteins (CKAMP52, -59, -39 and -44; Shisa6, -7, -8 and -9) (Jacobi & von Engelhardt, 2017; 2021), two proline-rich transmembrane proteins (PRRT1 and -2) (Schwenk et al. 2012(Schwenk et al. , 2014Shanks et al. 2014;Matt et al. 2018), and the leucine-rich repeat transmembrane neuronal protein 4 (LRRTM4) (Schwenk et al. 2014). Recent work has also shown that different auxiliary proteins associate with the AMPAR subunits during their assembly within the endoplasmic reticulum (ER). FRRS1l (ferric chelate reductase 1 like) protein, in complex with CPT1c (carnitine O-palmitoyl-transferase 1c), mediate the formation of GluA tetramers from monomers initially associated with ABHD6 (α/β-hydrolase domain-containing 6), and allow their co-assembly with the core AMPAR auxiliary subunits in readiness for exit from the ER and subsequent insertion in the plasma membrane (Schwenk et al. 2019). Many of the auxiliary subunits have been shown to modify basic properties of both CI-and CP-forms of AMPARs (see Table 1). Thus, the TARPs typically increase single-channel conductance, slow the channel kinetics, alter the pharmacology of agonists, antagonists and allosteric modulators, and enhance receptor trafficking to the cell surface (Jackson & Nicoll, 2011b;Greger et al. 2017;Jacobi & von Engelhardt, 2021). The degree to which they influence the AMPAR properties varies between TARP sub-family members. For example, type Ib TARPs (γ 4 and γ 8) slow the channel kinetics and can increase single-channel conductance to a greater extent than type Ia (γ 2 and γ 3) or type II (γ 5 and γ 7) TARPs (Cho et al. 2007;Milstein et al. 2007;Kato et al. 2010;Jackson et al. 2011).
In the case of GluA2-lacking CP-AMPAR, co-assembly with TARPs or CNIHs increases their already high (relative to CI-AMPARs) single-channel conductance (Tomita et al. 2005;Soto et al. 2009;Coombs et al. 2012) (see Fig. 2). For homomeric GluA4 receptors, it is thought that this increase reflects an enhanced proportion of events opening to their higher sub-conductance states (Tomita et al. 2005). By contrast, for GluA1 receptors there appears to be an increase in the absolute amplitude of the maximum conductance state . In all cases, the unusually high single-channel conductance of TARP associated CP-AMPARs is often sufficient to allow these to be distinguished from the TARPed CI-AMPARs, or indeed from TARPless AMPARs (see Bats et al. 2012). CNIHs and TARPs increase conductance to a similar extent (Coombs et al. 2012), while CKAMPs/Shisas produce only a marginal increase in channel conductance (Jacobi & von Engelhardt, 2017). In striking contrast with the other core auxiliary subunits, GSG1L reduces both the weighted mean single-channel conductance (by ∼50%) and the calcium permeability of CP-AMPARs, while increasing the channel's polyamine-dependent rectification (Fig. 2). Thus, increased expression of GSG1L has been found to reduce EPSC amplitude (McGee et al. 2015;Gu et al. 2016).
Co-assembly of CP-AMPARs with TARPs or CNIHs partially relieves the block by intracellular polyamines (Cho et al. 2007;Soto et al. 2007;Coombs et al. 2012;Brown et al. 2018), by increasing polyamine permeation (Brown et al. 2018). By contrast, TARPs enhance CP-AMPAR block by extracellular polyamine toxin PhTx-433 (Jackson et al. 2011). This block is more effective when the receptors are activated by the full agonist glutamate rather than by the partial agonist kainate, suggesting that the block is favoured when the channels open predominantly to higher conductances (Jackson et al. 2011). Indeed, the degree of block of CI-AMPARs by extracellular PhTx-74, a related polyamine toxin, is positively correlated with their single-channel conductance (Jackson et al. 2011). While a detailed mechanism for this observation is lacking, the idea that TARP-increased channel conductance and altered polyamine block might originate from a simple increase in the pore size can be excluded, as functional J Physiol 599.10 evidence suggests that the CP-AMPAR channel pore diameter is unaltered by TARPs (Soto et al. 2014).
Recent cryo-EM work has solved the structures of γ 2-associated CP-AMPARs (homomeric unedited GluA2) in the presence of the exogenous channel blockers NASPM, IEM-1460 and argiotoxin-636 (Twomey et al. 2018). Each blocking molecule sits along the pore axis of the channel with its hydrophobic head below the channel's gate and above the selectivity filter. The hydrophobic head stops the molecule from readily permeating through the channel, and the tail extends down through the selectivity filter. For all three blocking molecules the channel's Q/R site glutamines, which form the narrowing constriction of the pore, appear to be the main anchoring point for their tail. It is therefore suggested that the blockers supress current flow by plugging the ion channel, without interfering with the gating mechanism (Twomey et al. 2018). For intracellular polyamines, in addition to the Q/R site, electronegative charge provided by an aspartate residue at the 'Q/R +4' site is a key determinant of block. Neutralization of this charge decreases spermine block (Panchenko et al. 1999;Soto et al. 2014) as well as reducing channel conductance (Soto et al. 2014).
Co-assembly with TARP family members produces another surprising change in AMPAR pharmacology, transforming the competitive antagonist CNQX into a partial agonist, and increasing the efficacy of the partial agonist kainate (Jackson & Nicoll, 2011b). Of note, not all TARPs render AMPARs sensitive to activation by CNQX. The type II TARP γ 7 is ineffective in this respect (Bats et al. 2012), although it is still capable of relieving intracellular polyamine block and increasing channel conductance. Interestingly, CNIHs fail to convert CNQX to a partial agonist, and only marginally increase the efficacy of the partial agonist kainate (Shi et al. 2010).
TARPs also reduce AMPAR desensitization and enhance the efficacy of glutamate at the concentrations that prevail during fast transmission (Cho et al. 2007;Milstein et al. 2007;Ben-Yaacov et al. 2017;Coombs et al. 2017). Recently, we examined the influence of TARPs on AMPARs gated by low concentrations of glutamate in order to obtain information about receptor activation during slower and more diffuse synaptic events. By first saturating the receptors with the antagonist NBQX, then rapidly switching into glutamate, it was possible to observe directly the sequential gating responses as individual molecules of the competitive blocker slowly unbound to be replaced by glutamate. This provided information about the time course of channel activation (over hundreds of milliseconds) and revealed the sub-conductance level associated with each occupancy state of an individual TARPed receptor. Unlike TARPless receptors examined under similar conditions, that were found to exhibit three open levels, for TARPed CP-AMPARs, four directly resolved conductance steps were evident during the channel activation process. This indicates an enhancement of glutamate efficacy such that even singly liganded receptors are able to generate channel openings. While the single-channel conductance of such events is relatively small, ∼10% of the fully open state (Coombs et al. 2017), the overall effect of TARPs on glutamate efficacy will facilitate synaptic signalling and Ca 2+ influx (for CP-AMPARs) during prolonged exposure to low transmitter concentrations. This is likely to enhance AMPAR responses such as those that occur during synaptic spillover and delayed clearance of transmitter (DiGregorio et al. 2007;Zampini et al. 2016).

Normal and maladaptive forms of plasticity involving CP-AMPARs
A rapid and lasting change in the prevalence of GluA2-containing AMPARs, and thus in the fraction of CP-AMPARs, is a key feature of many different forms of synaptic plasticity (see Table 2). We will briefly summarize some of these, before focusing (in the section on: Auxiliary subunits implicated in native CP-AMPAR regulation) on those forms where information is available about the involvement of auxiliary subunits in this regulation.
As first described at cerebellar parallel fibre-to-stellate cell synapses, where a proportion of synaptic AMPARs are calcium-permeable, high frequency activity can trigger a change in the current-voltage (I-V) relationship of the EPSCs. This effect can be replicated by activation of synaptically located metabotropic (mGluR1) receptors (Fig. 3). Decreased CP-AMPAR expression following activation of mGluR1s is a theme in several different neuron types and of particular interest given its relevance  to drug addiction (see below). The rapid alteration from inwardly rectifying to linear I-V is accompanied by a reduction in EPSC amplitude (at negative potentials), reflecting the replacement of the CP-AMPARs by lower conductance CI-AMPARs (Liu & Cull-Candy, 2000;Gardner et al. 2005;Kelly et al. 2009). The activation of both CP-AMPARs and mGluR1/5 is necessary to trigger a rise in intracellular Ca 2+ required for this AMPAR plasticity, implying the presence of a self-regulating mechanism (Kelly et al. 2009;Liu et al. 2010;Bats et al. 2012;. Conversely, plasticity involving a lasting increase in CP-AMPAR expression appears to underlie several forms of synaptic remodelling that are physiologically and behaviourally important. These include postsynaptic changes in lamina ll spinal cord neurons as a result of inflammatory pain (Kopach et al. 2011;Sullivan et al. 2017) and synaptic remodelling associated with fear conditioning and fear extinction (Clem & Huganir 2010;Liu et al. 2010). Many detrimental types of plasticity have been described that involve an increase in CP-AMPAR expression. These include the cocaine-induced modification of glutamatergic transmission onto dopamine neurons in the ventral tegmental area and nucleus accumbens (Bellone & Luscher, 2006;Selvakumar et al. 2014), anoxia-induced decreases in GluA2 expression in hippocampal CA1 cells (Noh et al. 2005;Quintana et al. 2015), and increased CP-AMPAR expression in oligodendrocyte lineage cells that can follow hypoxia during gestation (Follett et al. 2004;Zonouzi et al. 2011;Ceprian & Fulton, 2019). Additionally, certain mutations in prion proteins can result in disorders that involve excitotoxic neurodegeneration caused by increased expression of neuronal CP-AMPARs (Ghirardini et al. 2020). And in a mouse model of glaucoma, elevated intraocular pressure causes an increase in damaging CP-AMPAR expression in specific subpopulations of retinal ganglion cells (Sladek & Nawy, 2020).
Although not a conventional plasticity, it is also interesting to note that the aberrant expression of CP-AMPARs can result from the downregulation of mRNA editing at the Q/R site of GluA2 (Wright & Vissel, 2012;Slotkin & Nishikura, 2013). This has been suggested to play a role in Alzheimer's disease (Gaisler-Salomon et al. 2014), in both sporadic and familial amyotrophic lateral sclerosis J Physiol 599.10 (motor neuron disease) (Yamashita & Kwak, 2019), in seizure vulnerability (Konen et al. 2020), and in the growth of malignant gliomas (Maas et al. 2001).

Auxiliary subunits implicated in native CP-AMPAR regulation
As various auxiliary proteins, including TARPs, CNIHs and GSG1L, can modify the biophysical behaviour and pharmacology of both CP-and CI-recombinant AMPARs, the question arises, are there specific auxiliary proteins that selectively regulate the trafficking and localization of native CP-AMPARs? To date, the auxiliary subunits involved in CP-AMPAR regulation and plasticity have been examined in only a relatively small number of cell types, but these studies have started to throw some light on this issue (see Table 3).

CP-AMPARs in glia, oligodendrocyte precursor cells and glioma.
Unlike most neurons Bergmann glia (BG), the main radial glia within the cerebellum, appear to be entirely devoid of GluA2 and thus express exclusively CP-AMPARs. The activation of these receptors by glutamate is crucial for BG cell processes to correctly ensheath the synapses present on Purkinje cell dendritic spines, thereby enabling fast transmission, transmitter removal, and optimal synaptic integration (Iino et al. 2001;Saab et al. 2012). BG strongly express γ 5, a TARP that is absent from all other cerebellar cells (Fukaya et al. 2005).
The CP-AMPAR-mediated quantal events underlying neuron-glia signalling in BG display single-channel and kinetic properties indicative of γ 5-associated receptors, and there is good evidence to suggest the receptors are assembled from GluA1/γ 5 and/or GluA4/γ 5 (Soto et al. 2009). However, it is notable that BG also express γ 7 (Yamazaki et al. 2010). It thus remains possible that the functional properties of CP-AMPARs in BG cells are regulated by both γ 5 and γ 7. It is of particular interest that BG, which are unusual in expressing only CP-AMPARs, express only type II TARPs. This strongly suggests that, in some cell types at least, type II TARPs are sufficient to deliver CP-AMPARs to the plasma membrane. CP-AMPARs also play an important role in oligodendrocyte precursor cell (OPC) proliferation, differentiation, migration and neuron-glial signalling (Harlow et al. 2015;Chen et al. 2018). However, they also render OPCs particularly susceptible to damage during gestation and early stages of development (Volpe, 2009). We have identified several factors that regulate the AMPAR subtypes present in OPCs. Notably, activation of group 1 mGluRs in these cells triggers an increase in the proportion of CP-AMPARs, signified by an increase in inward rectification of glutamate-evoked currents. Furthermore, the kinetic features and underlying channel conductance of the CP-AMPAR-mediated currents suggest that these are TARP-associated AMPARs. Oligodendrocyte lineage cells express predominantly GluA2, -3 and -4 subunits, although GluA3 and -4 may predominate (Zonouzi et al. 2011;Zhang et al. 2014). Of these, GluA4 is thought to be particularly important in generating excitotoxic damage (Begum et al. 2018). TARPs γ 2, -3, -4 and -5 have all been detected in OPCs using RT-PCR (Zonouzi et al. 2011).
In addition, antibody labelling in these cells has verified the presence of TARPs that contain the TTPV motif at their C-terminus (Zonouzi et al. 2011), indicating that the predominant forms present are γ 2, -3 or -4 (rather than γ 5). Transfection of OPCs with a form of γ 2 that lacked its last 16 residues (including the TTPV motif required for binding to PDZ domain-containing proteins) was able to inhibit the mGluR-mediated increase in CP-AMPARs, leaving a glutamate-evoked current that was mediated entirely by CI-AMPARs. This confirmed the importance of type I TARPs in delivery of CP-AMPARs in these cells (Zonouzi et al. 2011). Interestingly, while there is also evidence that CNIHs are associated with AMPARs in OPCs (Coombs et al. 2012), in contrast with the TARPs there is no evidence to suggest CNIHs selectively target CP-AMPARs in OPCs. Thus, γ 2 appears to be the primary candidate for CP-AMPAR trafficking and localization in the OPC plasma membrane. AMPAR signalling is crucial for myelination but seems to enhance oligodendrocyte survival rather than promote myelination itself (Kougioumtzidou et al. 2017). Furthermore, OPC proliferation and differentiation are promoted differently, depending on the subunit composition of the AMPARs that are activated by axonal glutamate. OPC proliferation is triggered by expression of unedited CP-AMPARs, whereas the presence of GluA2-containing receptors appears to be required for OPCs to respond to differentiation cues (Chen et al. 2018). Clearly, understanding the mechanism by which TARPs (probably γ 2) target different AMPAR subtypes to influence OPC proliferation and survival could be of considerable therapeutic value.
Gliomas (oligodendrogial or astrocytic primary brain tumours) strongly express CP-AMPARs. These receptors assemble primarily from GluA1, -2 and -4. However, the presence of editing deficient forms of GluA2 within glioma cells (Maas et al. 2001;Venkataramani et al. 2019;Venkatesh et al. 2019) means that a high proportion of the AMPARs are highly Ca 2+ permeable despite the incorporation of GluA2. It has recently been shown that CP-AMPARs are present at synapses that form between neurons and glioma cells within the tumour, and that their activation promotes tumour cell proliferation and invasiveness. Suppression of activation by genetically perturbing AMPAR signalling with a dominant negative AMPAR subunit, or by the use of AMPAR antagonists such as parampanel (Venkataramani et al. 2019;Venkatesh et al. 2019) greatly reduces cell proliferation -a feature that has clear therapeutic potential. While little is currently known about the core auxiliary subunits involved in delivery of CP-AMPARs at these neuron-glioma synapses, it has long been known that the AMPAR-associated protein CPT1c is common in gliomas and in a surprisingly wide variety of other cancer cell types. These include lung, breast and pancreatic cells (reviewed in Chen et al. 2020). Several recent studies have highlighted the importance of CPT1c in AMPAR biogenesis and shown that it forms an integral part of the AMPAR complex in healthy cells throughout the CNS (Schwenk et al. 2012). While it clearly behaves as an interacting protein in heterologous expression systems (Gratacòs-Batlle et al. 2015) it does not appear to modify the functional properties of AMPARs.
Within neurons it does not associate with the AMPARs present in the plasma membrane, rather it forms part of the AMPAR assembly within the ER membrane, where it is crucial in tetramerization of the receptor dimers (Schwenk et al. 2019). Thus, for reasons that are far from clear, many tumour cell types including ones not associated with the nervous system, express AMPARs (and hence CPT1c) that appear to play a role in cell proliferation. This has been utilised as a novel marker of cancer cells as well as a potential therapeutic target that can be supressed (Zhang et al. 2017).
Acidosis/hypoxia in hippocampal CA1 region. Pyramidal cells in the CA1 region of the hippocampus are susceptible to damage following ischaemic stroke, where oxygen/glucose deprivation (OGD) promotes excessive glutamate release and acidosis that causes Ca 2+ influx. This triggers various downstream effects, including an increase in CP-AMPARs, activation of which allows a further rise in intracellular Ca 2+ that contributes to the delayed neuronal death (Opitz et al. 2000;Noh et al. 2005;Quintana et al. 2015). The shift in AMPAR subtype involves the rapid and selective endocytosis and lysosomal degradation of GluA2/GluA3 heteromers, a down-regulation of GluA2 transcription, and the recruitment of extrasynaptic CP-AMPARs (GluA1/GluA3 or homomeric GluA1) (Koszegi et al. 2017).
Of note, a transient recruitment of GluA1-containing CP-AMPARs to CA1 synapses has also been proposed to play a role during conventional long-term potentiation (LTP) and long-term depression (LTD) (Plant et al. 2006;Lu et al. 2007;Guire et al. 2008;Sanderson et al. 2016;Park et al. 2019). However, with regard to LTP, there is also evidence against recruitment of CP-AMPARs (Adesnik & Nicoll, 2007;Grey et al. 2007;Granger et al. 2013), and the topic remains unresolved (Purkey & Dell' Acqua, 2020). Roles for γ 8 and γ 2 have been proposed in LTP and LTD at CA1 synapses (Matsuda et al. 2013;Park et al. 2016;Sheng et al. 2018), but their interaction with CP-AMPAR subtypes has not been examined. Likewise, there is no clear indication of which auxiliary subunits are involved in delivery of CP-AMPARs following anoxia in CA1. TARPs γ 2 and γ 8, GSG1L, CNIH2 and CNIH3 are all present, and thus all are potential candidates. Interestingly, GSG1L has been shown to supress CP-AMPAR function and 'negatively regulate' synaptic transmission. Hence, GSG1L attenuates single-channel conductance and calcium permeability of homomeric AMPARs but increases block by intracellular spermine and increases mEPSC rectification in cultured cerebellar neurons (McGee et al. 2015). On the other hand, in hippocampal pyramidal cells knockdown or knockout of GSG1L enhances AMPAR-mediated synaptic transmission (McGee et al. 2015;Gu et al. 2016) and enhances LTP at the Schaffer-collateral pathway (Gu et al. 2016).

J Physiol 599.10
Addictive drug-induced changes in the ventral tegmental area and nucleus accumbens. Exposure to drugs of abuse causes various forms of synaptic plasticity within brain regions implicated in reward and motivation, notably the nucleus accumbens (NAc) and ventral tegmental area (VTA) (Luscher, 2016;Wolf, 2016). In dopamine neurons of the VTA that project to the NAc, a single exposure to cocaine, for example, alters excitatory transmission by promoting insertion of GluN3A-containing NMDARs triggering a subsequent switch from CIto CP-AMPARs and consequent potentiation of the synaptic currents (Bellone & Luscher, 2006;Yuan et al. 2013). An increase in the prevalence of CP-AMPARs is also seen in medium-spiny neurons of the NAc shell following withdrawal from cocaine (Conrad et al. 2008;Scheyer et al. 2018). In both cases, the increased neuronal excitation is thought to contribute to enhanced drug-related behaviours. While there is little information about the identity of auxiliary proteins involved in AMPAR changes in the VTA, biochemical studies using subcellular fractionation and antibody labelling in the NAc have suggested that the newly inserted synaptic GluA1-containing CP-AMPARs are associated with γ 2, and the extrasynaptic CP-AMPARs with γ 4 (Ferrario et al. 2011). It is interesting to note that γ 8 is also very abundant in NAc, while γ 7 and GSG1L are also present at a lower level (Schwenk et al. 2014, Supplementary Table). The involvement of these other potentially relevant auxiliary subunits is unknown. Subsequent studies have revealed an increase in both γ 2 and γ 4 in NAc following sensitization and withdrawal, and concluded that NMDAR-driven S-nitrosylation of γ 2, which increases GluA1/γ 2 association (Selvakumar et al. 2009), is necessary for the upregulation of surface GluA1-containing AMPARs (Selvakumar et al. 2014). Interestingly, in animals that have undergone incubation of cocaine craving, activation of mGlu1 receptors in the NAc triggers the endocytosis of the newly inserted CP-AMPARs (McCutcheon et al. 2011). As mGluR1 activation can also drive the synapse to its pre-drug state in VTA neurons, this lasting change has been suggested to offer a potential therapeutic target for reducing cue-induced craving (Bellone & Luscher, 2006;Scheyer et al. 2018).
Hyperalgesia in lamina ll of spinal cord. TARPs γ 2 and γ 8 are both present in lamina II of the superficial dorsal horn (SDH) of the spinal cord (Sullivan et al. 2017), an area involved in nociception. Heightened pain sensitivity associated with peripheral inflammation involves an increase in neuronal excitability and CP-AMPAR prevalence (Katano et al. 2008;Park et al. 2009). We have established that one of the mechanisms contributing to peripheral inflammation-associated changes is synaptic remodelling, characterised by an increase in CP-AMPARs specifically at the pain fibre synapses (Sullivan et al. 2017). Prior to hyperalgesia, transmission from local inputs onto lamina II neurons is mediated by γ 2-associated CI-AMPARs, while at peripheral pain fibre (C-fibre) synapses on the same cells it is mediated by CI-AMPARs associated with a different auxiliary subunit (possibly γ 8). The view that γ 2 is 'synapse specific' and absent from normal C-fibre synapses prior to hyperalgesia is supported by evidence from immunohistochemical co-labelling. Interestingly, the inflammation induced remodelling of C-fibre synapses entails replacement of the γ 2-lacking CI-AMPARs with γ 2-containing CP-AMPARs -a change that predictably is lost in the stargazer mouse (Sullivan et al. 2017), a mutant devoid of functional γ 2.
Recent work has identified a number of AMPAR antagonists that are highly selective for γ 8-associated receptors Maher et al. 2016;see below). One of these, LY3130481, has been shown to supress excitatory postsynaptic transmission and attenuate short-term synaptic plasticity in spinal sensory neurons, and supress behaviour associated with pain perception (Knopp et al. 2019). Although the precise role of γ 8 in spinal cord pain pathways is still uncertain, this evidence suggests that γ 8-selective antagonists could offer novel therapies for conditions involving chronic pain. Of note, LY3130481 only partially supresses EPSPs in dorsal horn spinal cord neurons, in contrast with the full block produced by the non-selective AMPAR antagonist GYKI53784. This may reflect the degree of γ 8 expression but also the relative expression of other TARPs, specifically γ 2. When tested against recombinant receptors, the potency and efficacy of LY3130481 is decreased by co-expression of γ 2 (or γ 3) with γ 8 (Knopp et al. 2019). The precise interplay of γ 8 and γ 2 in spinal nociceptive signalling remains to be determined. One possibility from earlier work (Sullivan et al. 2017) is that γ 8, along with γ 2, is required for the translocation of CP-AMPAR at the C-fibre synapses following peripheral inflammation. In which case, pharmacological block of γ 8-associated receptors could provide a promising approach for suppressing the inflammatory pain-induced plasticity. Cerebellar parallel fibre-stellate cell synapses. Cerebellar stellate cells (SCs) normally express CP-AMPARs at their parallel fibre inputs from granule cell (GCs). High frequency presynaptic activity triggers a rapid switch from CP-AMPARs to GluA2-containing CI-AMPARs causing a postsynaptic form of LTD (Liu & Cull-Candy, 2000;Gardner et al. 2005), a change that can also be generated by activating mGluR1 receptors with an applied agonist. Furthermore, experiments using mGluR1 blockers have demonstrated that tonic mGluR1 activation normally exerts a suppressive effect on CP-AMPAR expression at these synapses (Kelly et al. 2009). This has provided a useful model for comparison with other forms of CP-AMPAR plasticity, including those that are detrimental. Indeed, SC plasticity shows some intriguing parallels to that at synapses in the VTA where similar changes occur developmentally, even though the AMPAR subunits and auxiliary proteins are likely to differ (Mameli et al. 2011;Loweth et al. 2013).
Experiments on SC synapses have also allowed a direct test of whether γ 2 is required for CP-AMPARs to localize at synapses in the cerebellum. GCs and SCs each express two TARPs, γ 2 and -7. In the stargazer mouse, where SCs express only TARP γ 7, parallel fibre stimulation still evokes EPSCs (Bats et al. 2012). However, these were shown to be readily blocked by PhTx-433, indicating that the current was carried by CP-AMPARs. Although the CP-AMPARs could still localise at synapses in the absence of γ 2, they were strongly inwardly rectifying due to block by intracellular polyamines. This, together with a low single-channel conductance and slow kinetics, suggested the synaptic CP-AMPARs were TARPless, while the extrasynaptic ones had characteristics of TARPed CP-AMPARs. γ 7 is the only remaining TARP in stargazer SCs, suggesting that although synaptic receptors are likely to be TARPless, those in the extrasynaptic membrane are associated with γ 7. A different study also observed a dramatic increase in rectification of SC EPSCs (and of extrasynaptic AMPARs) in stargazer mice (Jackson & Nicoll, 2011a). However, as the authors were unable to detect any increased sensitivity to block by PhTx-433, they concluded that the increased rectification was unlikely to result from a decrease in AMPAR GluA2 content and hypothesized that it may be attributable to a TARP-dependent change in receptor gating. Thus, while both these studies suggest a role for γ 7 in AMPAR trafficking in stargazer SC cells, one concluded it acts non-selectively (Jackson & Nicoll, 2011a) while the other suggests it more likely promotes the presence of synaptic CP-AMPARs by normally suppressing synaptic expression of CI-AMPARs while allowing CP-AMPARs to localize at synapses (Bats et al. 2012).
The principle that CP-AMPARs can localize at central synapses in the absence of γ 2 has been tested more generally by examining synapses in cerebellar GCs from stargazer mice. These are devoid of miniature EPSCs (Hashimoto et al. 1999;Tomita et al. 2005), offering an unequivocal experimental scenario. GCs do not normally express CP-AMPARs, but when GluA2 was knocked down using siRNA mEPSCs unexpectedly reappeared (Studniarczyk et al. 2013). These currents were strongly inwardly rectifying suggesting that CP-AMPARs can indeed localize at synapses in the absence of γ 2 and the presence of γ 7. Furthermore, transfecting γ 7 into wild type GCs (which normally express only CI-AMPARs) gave rise to inwardly rectifying mEPSCs and whole-cell currents, supporting the view that γ 7 actively enhances CP-AMPAR expression (Studniarczyk et al. 2013). In contrast to these findings, experiments using a knockout mouse have suggested that γ 7 does not make a significant contribution to excitatory transmission in either cerebellar SCs or GCs (Yamazaki et al. 2015). Thus, at present, the possible role of TARP γ 7 in determining features of CP-AMPAR transmission remains unresolved.
Overall, it is clear that TARPs γ 2, -5, -7 and -8 and the atypical auxiliary subunit GSG1L are all potential 'molecules of interest' in the regulation of CP-AMPARs.

Possible pharmacological and molecular interventions
Changes in the regulation and function of CP-AMPARs occur in a wide variety neurological conditions and chronic disorders. Therefore, interventions that prevent the aberrant expression, trafficking or targeting of these receptors, or selectively reduce their damaging activation, could prove highly beneficial. A more complete understanding of the molecular mechanisms that underlie CP-AMPAR regulation is a crucial first step. In this respect an interesting theme has emerged from work that has shown CP-AMPAR expression to be decreased by the activation of mGluR1s in several different neuron types. Thus, enhancing mGluR1 activation using positive allosteric modulators, which has been suggested as a possible strategy for reversing increased CP-AMPARs associated with use of cocaine and other addictive drugs (Scheyer et al. 2018;Wolf, 2016), could have wide potential.
Epilepsy is another case where insight into the details of CP-AMPAR regulation has proved useful. One of the notable changes that follows seizures in humans and in mouse models of epilepsy is a dramatic increase in the expression of flip isoforms of GluA1. These not only confer greater glutamate sensetivity than the flop isoforms they replace, but if present in excess could form homomeric CP-AMPARs. Either of these features might be expected to enhance excitatory synaptic currents. A study by Lykens et al. (2017) reported the development of a splice modulating oligonucleotide that decreased GluA1 expression and showed anti-seizure properties, including reduced post-seizure hyperexcitability in neonatal mice. Such targeting of specific AMPAR subunit isoforms may have the potential for altering the expression of AMPAR subtypes involved other disease states. Likewise, various molecular approaches, including the use of small interfering peptides (Fosgerau & Hoffmann, 2015), have been used successfully to target protein-protein interactions and prevent the endocytosis of AMPARs involved in behavioural sensitization models of drug addiction (Dias et al. 2012). Small interfering peptides have also been developed to selectively prevent endocytosis of AMPARs containing GluA2 subunits (Lin et al. 2016). Clearly, it would be of interest to further develop such approaches to target specific auxiliary subunits that may be involved in CP-AMPAR delivery. J Physiol 599.10 Perhaps more immediately tractable is the goal of selectively modifying the function rather than expression of CP-AMPARs. Although the potential of AMPARs as therapeutic targets has been long recognised (see, for example, Bowie, 2008;Rogawski, 2011;Chang et al. 2012) there are unique challenges in attempting to pharmacologically interfere with a receptor that is both widespread in the CNS, and fundamental to most aspects of normal brain function. The novel properties of GluA2-lacking CP-AMPARs mean that, experimentally at least, it is possible to selectively block their integral ion channel (with existing pharmacological tools such as PhTx-433, IEM-1460 and NASPM). These blockers have helped reveal the surprisingly widespread involvement of CP-AMPARs in various forms of plasticity, including those contributing to neurological disease, and shown diverse therapeutic use in many preclinical studies (e.g. Noh et al. 2005;Yennawar et al. 2019;Hu et al. 2020;Adotevi et al. 2020). As yet, analogues of these drugs have not been successfully developed for wider use, but elegant cryo-EM work has recently provided invaluable insight into the architecture of the blocker binding site within the pore, and this is likely to provide significant impetus to the further development of small molecule blockers (Twomey et al. 2018).
The value of region-specific therapeutic intervention that can be gained by identifying molecules that target receptor-associated auxiliary proteins has been considered in recent reviews (Maher et al. 2017;Kato & Witkin, 2018). Several such molecules have been described that act as selective antagonists for AMPARs associated with γ 8, notably JNJ-55511118 (Maher et al. 2016), LY3130481/CERC-611 Kato et al. 2016) and JNJ-61432059 (Savall et al. 2019). These negative allosteric modulators appear to functionally disrupt the interaction between γ 8 and the pore-forming subunits in the AMPAR assembly and have shown promise as therapeutic approaches for epilepsy Savall et al. 2019) and chronic pain (Knopp et al. 2019). The binding of these molecules depends on specific amino acid residues within transmembrane regions of γ 8. Introducing the same residues into γ 2 and -4 confers drug sensitivity on receptors containing these modified TARPs (Maher et al. 2016). Of note, a recent study, using molecular dynamics simulations and electrophysiology revealed a conserved moiety among structurally diverse compounds that underlies their interaction within the binding pocket of γ 8 (Dohrke et al. 2020). The discovery of γ 8-selective drugs, and the growing understanding of how these may act, are exciting developments that could pave the way to the design of antagonists selective for AMPARs containing other TARPs, raising the prospect of tools for region-specific and CP-AMPAR subtype-selective intervention within the CNS.