Transcriptomic Expression of AMPA Receptor Subunits and their Auxiliary Proteins in the Human Brain

Abstract

Receptors to glutamate of the AMPA type (AMPARs) serve as the major gates of excitation in the human brain, where they participate in fundamental processes underlying perception, cognition and movement. Due to their central role in brain function, dysregulation of these receptors has been implicated in neuropathological states associated with a large variety of diseases that manifest with abnormal behaviors. The participation of functional abnormalities of AMPARs in brain disorders is strongly supported by genomic, transcriptomic and proteomic studies. Most of these studies have focused on the expression and function of the subunits that make up the channel and define AMPARs (GRIA1-GRIA4), as well of some accessory proteins. However, it is increasingly evident that native AMPARs are composed of a complex array of accessory proteins that regulate their trafficking, localization, kinetics and pharmacology, and a better understanding of the diversity and regional expression of these accessory proteins is largely needed. In this review we will provide an update on the state of current knowledge of AMPA receptors subunits in the context of their accessory proteins at the transcriptome level. We also summarize the regional expression in the human brain and its correlation with the channel forming subunits. Finally, we discuss some of the current limitations of transcriptomic analysis and propose potential ways to overcome them.

1. Introduction

Glutamate receptors of the AMPA type (AMPARs) are the principal drivers of fast excitatory transmission in the brain, and in concert with receptors to GABA, the principal inhibitory neurotransmitter, establish the synaptic Excitation to Inhibition balance which is fundamental for proper brain function [19, 88]. Therefore, disturbances in AMPARs have been linked to numerous psychiatric, neurodevelopmental, and neurodegenerative disorders including epilepsy, schizophrenia, autism, and Alzheimer’s disease [8, 24, 26, 78, 87]. The involvement of AMPARs in disease has driven strong research efforts that have greatly advanced our understanding of these receptors; yet there is only one FDA-approved drug whose mechanism of action is through the modulation of AMPARs. Perampanel is an AMPAR antagonist indicated as an adjuvant drug for refractory epilepsy [25] and its success indicates that modulation of AMPARs is a viable and effective strategy for correcting abnormal neuronal activity. However, perampanel effects on the ubiquitously expressed AMPARs seem to be responsible for some of the adverse effects, like gait disturbance and aggressive behavior, observed in clinical practice [124]. These observations have prompted the search for modulators of AMPARs with a more regionalized effect. The discovery of LY3130481 [51] or JNJ-55511118 [68] is a step in that direction. These drugs are anticonvulsants with potent antagonist activity on AMPAR complexes containing the auxiliary subunit TARP-γ8 (transmembrane AMPAR regulatory protein gamma 8, gene name: CACNG8). Because TARP-γ8 is largely expressed in the forebrain, especially in the hippocampus, but not in the cerebellum, it could have therapeutic effects in diseases characterized by hippocampal hyperexcitability without the adverse effects associated with the cerebellum such as motor control [51, 68, 129]. Proteomic studies have identified a plethora of auxiliary subunits of AMPARs which are anatomically isolated in discrete brain regions and have modulatory roles on the physiology of AMPARs [94]. Further, functional studies have confirmed that changes in TARP stoichiometry, particularly in the hippocampus, affect AMPAR gating and kainate efficacy [75, 102]. It is here, where a large gap of knowledge in the regional expression of AMPARs in humans under even healthy conditions, is present and where transcriptomics provides a golden opportunity to start filling this gap.

2. Transcriptomic diversity of AMPA receptor complexes

AMPA receptors are heteromeric dimers-of-dimers that draw from a pool of four subunits (GluA1-GluA4) encoded by four genes (GRIA1-4) [66, 110, 126]. The exact composition of these pore-forming subunits appears to be input-specific [30, 34, 35, 41, 44, 89, 112], cell-specific [34, 35], and activity-dependent [61, 62]. A larger functional heterogeneity arises from diverse heteromeric assemblies, alternative splicing [105], RNA editing [119] and post-translational modifications [67]. For instance, the presence of the GluA2 subunit participates in the calcium ion permeability of the receptor as receptors lacking this subunit exhibit increased calcium permeability and are sensitive to voltage-dependent polyamine modulation [7, 27, 40, 114]. Recoding of arginine to glycine (R/G) or glutamine to arginine (Q/R) in three of the four AMPAR pore forming subunits provide an additional point of regulation for calcium permeability and the kinetics of the receptor responses [65].

AMPARs also rely on a host of auxiliary proteins that interact directly or indirectly with the pore-forming subunits [10, 18]. The accessory proteins can be split into those composing a defined core together with GluA1-GluA4 subunits, and those located in the periphery of the core and assemble through direct binding or temporal interactions (Table 1). Accessory proteins in the core are thought to act as a bridge for the effects of the peripheral proteins on the pore forming subunits [94]. This core can then be further subdivided into inner core proteins that occupy two pairs of distinct binding sites in the complex and outer core components that bind to other interaction sites. These accessory proteins regulate assembly, trafficking, synaptic localization, anchoring, and receptor kinetics and pharmacology [1, 6, 17, 18, 42, 49]. The importance of studying AMPARs with their accessory proteins is clearly highlighted by the work of Menuz and colleagues [74]. CNQX (6-cyano-7-nitroquinoxaline-2,3-dione) is widely used as an antagonist of AMPARs [38]; indeed, in heterologous systems CNQX blocks the effect of agonists by competing for the same site [115]. However, when members of the TARP family are present in the AMPAR complex, CNQX acts instead as a partial agonist and will induce a fraction of the activation of the receptor [74]. Since most native AMPARs in the brain contain at least one member of the TARP family CNQX most likely acts as a partial agonist of AMPARs in vivo [74]. The direct activation of AMPAR in complex with TARP proteins by CNQX would explain why CNQX leads to an increased spiking activity of interneurons in the hippocampus [4, 32, 73], and neurons in the thalamic reticular nucleus [56] and locus coeruleus [86].

Table 1.

Accessory proteins of AMPA receptors

TYPE	COMPLEX	REGULATES	GENE NAME	REFERENCES
CORE	Inner Core	Pharmacology, kinetics, trafficking	CACNG2, CACNG3, CACNG4, CACNG8	[12, 48, 50, 52, 106, 111]
		Pharmacology, kinetics	CACNG5, CACNG7
		Surface expression & gating properties	CNIH2, CNIH3	[95]
		Trafficking & channel conductance	GSG1L	[94, 99]
	Outer Core	Complex assembly	FRRS1L, NRN1	[9, 31, 94]
		Trafficking & stabilization	PRRT1, PRRT2	[59, 94, 113]
		Receptor kinetics	SHISA9	[22]
PERIPHERY	Periphery	Uncertain	ABHD12, LRRTM4, MPP2, OLFM2, OLMF3, SACM1L, VWC2, VWC2L	[94]
		Lateral mobility	OLFM1	[83]
		Complex assembly	CPT1C, ATAD1, PORCN	[125]
		Regulates receptor activity	SHISA7, CAMK2A, CAMK2B, CAMK2D, CAMK2G	[23, 92]
		Signaling	GNAS	[43]
		Surface expression	ABHD6, MYO5B, MYO5A, NSF, SYT1, SYT7	[2, 3, 15, 79, 82, 117, 118, 120, 122]
		Homeostasis and Synaptic plasticity	GOPC	[16]
		Synaptic plasticity	NRP2, PLXNA3, AKAP5, MAGI2, AAK1, AP2A1, AP2A2, ACTN1, STX4, NAPB, CDH2, PICK1, RAB5A, RAP2B	[3, 5, 17, 47, 53, 57, 58, 72, 90, 109, 116, 128]
		Trafficking and clustering	KIF1A, KIF5B, PPFIA1, GRIPAP1, DLG1, DLG2, DLG3, DLG4, EPB41, EXOC4, PLXNA4, IQSEC1, STX12, SHISA6, GRIP1, GRIP2, AP4B1, AP4E1, AP4M1, AP4S1, SYNDIG1, SEMA3A	[3, 13, 21, 28, 39, 45, 55, 60, 69, 71, 77, 98, 101, 103, 116, 121, 123]

Based on these complexities, studying AMPA receptors by looking at individual subunits separately or in the absence of their accessory proteins provides limited access to the potential role of AMPARs in disease.

3. Regional expression of AMPA receptors and its auxiliary subunits

The AMPAR knowledge base is larger for the hippocampus than any other brain region. For example, a PubMed search with the terms “AMPA receptors x”, where x is hippocampus, cortex, cerebellum, or thalamus, provides a total of 4594, 3071, 980 and 320 articles, respectively, published between the years 1983 and 2021, indicating that our knowledge of the particular stoichiometry of receptors in this region is more abundant than other equally important regions. The efforts of the Allen Institute have provided the field with a mine of information about the gene expression across the whole human brain [33, 108] facilitating the 30,000 ft view of gene expression and additional patterns not observed by studying only few brain regions. This data has shown that the gene expression of the pore-forming subunits GRIA1-4 in the brain is highly stereotypical across healthy individuals [100], following a pattern of brain region specificity according to their embryonic origin [54]. Indeed, gene expression separation by unsupervised clustering analysis suggests that GRIA1-4 are strong gene markers of synaptic components found in bioinformatic analysis [46, 54]. Specifically, the GRIA1 subunit is highly expressed in the hippocampus and amygdala and is highly conserved across species [29]. The cerebellum exhibits the highest levels of GRIA4 expression while the hippocampus exhibits the lowest [100].

GRIA2 is the most expressed subunit in the brain, which is a logical consequence of it forming part of the three major populations of AMPARs complexes in the brain (e.g. GluA1/2, GluA2/3 and GluA1/2/3) [127]. Using subunit specific antibodies in combination with cryo-electron microscopy (Cryo-EM), Zhao and colleagues were able to map the subunit composition and spatial arrangement of the GluA2 containing native AMPARs, as well as the relative abundance of the distinct native AMPARs compositions. They described 10 complexes, representative of native AMPARs across the whole brain, noting that the spatial arrangement of native AMPARs in specific brain regions is still missing and will require further studies [127]. Importantly, the structure of the most abundant AMPAR in the brain, GluA2/3, differs between the native complex analyzed by cryo-EM and the one analyzed in recombinant studies [36], thus highlighting once again the importance in including accessory proteins when studying the receptor complexes.

The recent development of the first positron emission tomography (PET) tracer to visualize AMPARs is a new tool to validate or reevaluate previous results from transcriptomics analysis. A derivative of 4-[2-(phenylsulfonylamino)ethylthio]-2,6-difluoro-phenoxyacetamide radiolabeled with ¹¹C ([¹¹C]K-2) recently showed specific binding to AMPARs [76]. This advancement has allowed for the first time the visualization of these receptors in live humans. Interestingly, [¹¹C]K-2 binding to hippocampus and amygdala was lower than the cortex, which in turn was also lower than cerebellum [76], suggesting that the hippocampus has less AMPARs compared to other brain regions. At first glance, this contrasts with transcriptomic analysis that found the highest mRNA levels for AMPARs in the hippocampus [100]. However, [¹¹C]K-2 has a higher affinity for GluA2 and GluA4 and only modest affinity to GluA1 and GluA3 subunits. Therefore, the pharmacological profile of [¹¹C]K-2 do reflects the transcriptional levels of these subunits in the brain [100], with little binding of AMPARs in the hippocampus due to high GRIA1 expression and robust binding in the cerebellum due to high GRIA4 expression. Although the relationship between mRNA and proteins is gene specific and not necessarily linear [20], recent analysis of GABA_ARs in the human brain [97] found strong correlations between mRNA levels and benzodiazepine binding measured in PET studies of cognitively normal controls [80], indicating that transcriptomic analysis across the brain correlates with PET imaging studies in at least some instances. Further analyses also have shown that the abundance and relationships between transcripts for families of receptors can be very informative of the major receptor complexes previously unknown in specific cell types. For example, by using the fractional contribution of mRNA for GABA_ARs in oligodendrocyte precursor cells, which represents the available pool of GABA_AR subunits mRNA, the stoichiometry of the most expressed GABA_AR receptor in these cells can be deduced and functionally validated by electrophysiological and pharmacological studies [81]. Therefore, it is expected that the pharmacological and binding profile of [¹¹C]K-2 will be an important tool in research and clinical practice involving AMPA receptors in alive people. Indeed, recent [¹¹C]K-2 studies confirmed an increased number of AMPARs in mesial epilepsy which was initially suggested by early measurements of mRNA levels by in situ hybridization studies in surgically collected and autopsy tissue from patients with intractable temporal lobe seizures [70].

The regional expression of accessory proteins is well known in some areas like the hippocampus and cerebellum, but to the extent of our knowledge the regional expression across the whole brain has not yet been explored. Here, we analyzed the gene expression of AMPAR accessory proteins across 87 brain regions (Figure 1). It is clear that accessory proteins also follow a regional expression which overlaps to some extent with the differential expression of the GRIA1-4 subunits, although some show a divergent pattern. This regional expression pattern is further demonstrated using a covariance analysis which also shows that these divisions into major brain regions can be driven by specific genes (Figure 2). For instance, MPP3, EPB41 and LGALS12 completely separate the cerebellum from all other brain regions. Similarly, a combination of accessory proteins like CACNG4 and SHISA6 drives the separation of subcortical regions deriving from the diencephalon. Thus, the inclusion of accessory protein expression provides a more comprehensive method of defining brain regions than relying solely on the GRIA1-4 subunits.

Figure 1. Whole brain gene expression of AMPA receptor subunits and their accessory subunits.

Heat plot shows the average Log2 mRNA gene expression (n= 6 subjects) downloaded from the Allen Human Brain Atlas (https://human.brain-map.org), for the most representative probes for each gene selected by factorial analysis as previously reported (see Supplementary Table 1 for additional information about the selected probes) [96, 100]. GRIA1-4 are shown in red. Labels for brain substructures are colored as per the insets. FL frontal lobe, Ins insula, CgG cingulate gyrus, HiF hippocampal formation, PHG parahippocampal gyrus, OL occipital lobe, PL parietal lobe, TL temporal lobe, Amg amygdala, GP globus pallidus, Str striatum, Cl claustrum, Hy hypothalamus, SbT subthalamus, DT dorsal thalamus, VT ventral thalamus, MES mesencephalon, CbCx cerebellar cortex, CbN cerebellar nuclei, Bpons basal part of the pons, PTg pontine tegmentum, MY myelencephalon. For substructure abbreviations, please see Supplementary Table 2

Figure 2. Region-specific gene expression of AMPA receptor subunits and their accessory subunits in the human brain.

Principal component analysis on covariances using JMP v.14 shows the separation of brain regions based on the expression of AMPARs complexes. Major brain regions are coded as per the insert. Each individual point corresponds to a single structure using data shown in Figure 1.

Past proteomics experiments in rats have established region-specific expression among the auxiliary proteins [93]. While the expression of AMPAR subunits and inner core genes aligned closely with the proteomics data, the expression of many outer core and peripheral genes did not (Figure 1). For instance, CACNG2 and CACNG4 exhibited the highest expression among the TARPs, corresponding to the higher protein expression while CACNG5 had the lowest transcript and protein expression. On the other hand, the cornichons exhibited flipped expression patterns, with CNIH2 exhibiting higher expression than CNIH3, although they retained their regional patterns of high expression in the hippocampus. Similarly, PRRT2 exhibited lower levels of expression than PRRT1, which had the highest expression of all genes analyzed, despite PRRT2 protein levels being higher. Among the peripheral genes, many exhibited altered regional patterns, particularly in the hippocampus and cerebellum. These differences may be due to variation between rats and humans, possible effects of post-transcriptional or translational modifications, or regulatory mechanisms that affect complex assembly. Therefore, integration of proteomics data with transcriptomics from the same individual would be helpful for determining the causes for this difference.

4. Conclusion

Transcriptomics offers a high-throughput method of distilling complex protein aggregation into simple metrics that can be used to compare between different brain states.

However, like any other method, this approach has its limitations. For example, although surveying for splice variants is possible, the determination of splice variants across the whole brain has not been done yet. This is important as all four pore forming subunits have two potential isoforms termed ‘flip’ and ‘flop’ that exhibit different channel kinetics and trafficking [14, 65, 84, 85, 105] and the distribution of these isoforms for the GluR1-3 subunits has been demonstrated to vary between brain regions [11]. Further, deregulation of the editing mechanisms has been suggested to underlie neuropsychiatric disorders such as schizophrenia and bipolar disorder [104]. Therefore, future analysis of data that account for different isoforms and RNA editing from tools such as REDItools [63, 64] could prove useful.

An important, although still difficult, goal in the search to understand neurotransmitter receptors in health and disease is to have a single cell transcriptome Atlas for the distinct cellular types in the whole brain. Taking in consideration that 1) there are 75 transcriptomically distinct cell types just in the middle temporal gyrus [37], 2) the human cortex can be parcellated histologically in 48 Broadman areas (BA) from which the middle temporal gyrus is BA21 (areas 13–16 are not defined in humans)[107], 3) and the abundant subcortical nuclei with increased cellular diversity [91], the task may seem daunting. However, collaborative research and systematic approaches like those promoted by Allen Institute and other groups will likely make this possible.

Highlights.

• Therapeutic challenges and opportunities of modulating AMPA receptors accessory proteins

• Regional-specificity of AMPA receptor auxiliary subunit gene expression across the whole human bain.

• Transcriptomic and in vivo imaging studies of AMPA receptors