Gabapentinoid Drugs Reduce Neurotransmitter Release

Previous studies showed that gabapentinoid drugs reduce the release of excitatory neurotransmitters from neuronal tissues by acting at the α₂δ-1 binding site [reviewed in Dooley et al. (2007)]. The prevailing thought has been that these drugs act therapeutically to reduce excitability in spinal cord sensory circuits and in neocortex by subtly reducing excitatory neurotransmitter release at many synapses at once, and glutamate synapses in many regions have abundant α₂δ-1 protein (Taylor and Garrido, 2008).

The amount of α₂δ-1 protein at synapses is remarkably upregulated after neuronal injuries, both in the spinal cord dorsal horn (Luo et al., 2001, 2002; Li et al., 2004, 2006; Bauer et al., 2009; Boroujerdi et al., 2011) and in neocortex (Andresen et al., 2014; Prince et al., 2016; Luo et al., 2018). Interestingly, this upregulation only alters α₂δ-1 and not α₂δ-2, and the upregulated α₂δ-1 is of a different post-translational splice variant and different glycosylation pattern than endogenous α₂δ-1 in the spinal cord (Luo et al., 2001) and dorsal root ganglia (Lana et al., 2014). Electron microscopy has shown localization of α₂δ-1 protein both presynaptically and postsynaptically in rodent dorsal spinal cord (Bauer et al., 2009), with about twice as much immunoreactivity presynaptically as postsynaptically. Also, after neuropathic injury, α₂δ-1 increases only at presynaptic sites (Fig. 3).

Several studies show that gabapentin and pregabalin reduce the spontaneous rate of release of vesicles from glutamate synapses, measured by miniature excitatory synaptic currents (mEPSCs), which reflect the release of individual synaptic vesicles. Unlike synaptic responses triggered by action potentials, mEPSCs are insensitive to the sodium channel blocker tetrodotoxin, and the vesicles released spontaneously are regulated differently from the vesicle pool released by action potentials [reviewed in Ramirez and Kavalali (2011)]. For example, spontaneous release occurs from a separate pool of vesicles (Sara et al., 2005) and is regulated by different synaptic proteins (Groffen et al., 2010) than evoked release. A recent study (Ferron et al., 2018) shows that the presence of mature α₂δ-1 in cultured neurons promotes both synchronous and asynchronous neurotransmitter release.

Such mEPSCs recorded in dorsal spinal cord (or trigeminal nucleus) neurons typically have rates in the range of 0.2–1.0 Hz, and the rate is augmented in response to experimental peripheral nerve damage (Li et al., 2014a,b; Zhou and Luo, 2015; Alles et al., 2017; Chen et al., 2018), chemotherapy-induced allodynia (Chen et al., 2019), and allodynia from tolerance to repeated morphine treatment (Deng et al., 2019) or from artificial excess expression of α₂δ-1 genes (Zhou and Luo, 2014). In each of these studies, the drugs gabapentin or pregabalin normalize pathologically elevated rates of mEPSCs in rat or mouse neuronal tissue with synapses in spinal cord dorsal horn (Patel et al., 2000; Li et al., 2014b; Matsuzawa et al., 2014; Zhou and Luo, 2014, 2015; Park et al., 2016; Alles et al., 2017; Chen et al., 2018, 2019; Deng et al., 2019). Gabapentinoids reduced the rate of mEPSCs between glutamatergic neurons in rat entorhinal cortex (Cunningham et al., 2004), in neocortex neurons after cortical freeze lesions (Andresen et al., 2014; Lau et al., 2017), between neocortex and striatal neurons (Zhou et al., 2018), at glutamate neurons in hypothalamus of spontaneously hypertensive rats (Ma et al., 2018), at the mouse calyx of Held (Di Guilmi et al., 2011), and in corticostriatal synapses after prolonged prior stimulation of striatum (Nagai et al., 2019). Therefore, the most widely replicated effects of gabapentin and pregabalin at the cellular level are decreases in the rate of mEPSCs at excitatory synapses, particularly at synapses with pathologically enhanced activity. Interestingly, gabapentin had little effect on the rate of mEPSCs recorded in inhibitory neurons in dorsal spinal cord (Zhou and Luo, 2015; Alles et al., 2017).

Despite agreement that gabapentinoid drugs reduce the rate of spontaneous mEPSCs and also subtly reduce bulk neurotransmitter release from whole tissues, it is not clear whether these effects result from decreased current through presynaptic calcium channel α₁ subunits. In fact, one set of studies showed that overexpression of α₂δ-1 in cultured cortex neurons increases neurotransmitter release and synaptic localization of calcium channels but decreases presynaptic calcium influx (Hoppa et al., 2012, 2014).

Additional Binding Partners of α₂δ-1

Although α₂δ-1 was originally identified in association with voltage-gated calcium channels, it has subsequently been found to associate with proteins other than calcium channels based on several methods (see Fig. 2 and Table 1). There is also evidence from fluorescent microscopy of living cells (Table 2), including the extracellular matrix proteins, collagen, and thrombospondin. In addition, gabapentin binding to α₂δ-1 prevents both thrombospondin and NMDA receptors from augmenting the release of glutamate at synapses and also prevents thrombospondin from promoting the formation of new glutamate synapses (see sections below). The α₂δ-1 protein has several unusual properties that may contribute to its selectivity in interacting with other proteins. It is highly glycosylated (decorated with polysaccharides) and is found concentrated in cholesterol-rich and detergent-resistant membrane microdomains called lipid rafts (Davies et al., 2006; Dolphin, 2013, 2016). In addition, mature α₂δ-1 seems to exist in some situations bound to membranes by a glycosylphosphatidylinositol anchor (GPI anchor), attached during post-translational protein processing (Davies et al., 2010). The GPI anchor is needed for α₂δ-1 interaction with calcium channel α₁ (Davies et al., 2010) but not for interactions with thrombospondin (Risher et al., 2018) or NMDA receptors (Chen et al., 2018). The GPI anchor adheres the protein to only the outer leaflet of the plasma membrane (Alvarez-Laviada et al., 2014) and favors localization in cholesterol-rich membrane areas. The putative relationships of α₂δ-1 with several of these known protein binding partners are shown in Fig. 3. Three-dimensional structures of partner proteins have been determined, including the voltage-gated calcium channel (Wu et al., 2016), α-neurexin (Miller et al., 2011), thrombospondin (Carlson et al., 2008), heteromeric NMDA receptors (Karakas and Furukawa, 2014), and BK calcium-dependent potassium channels (Yuan et al., 2010). The evidence for gabapentinoid drugs acting via these various proteins is discussed below, beginning with NMDA receptors, for which there is the most evidence.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Proposed direct molecular interactions of α₂δ-1 protein (yellow) with other neuronal proteins. The outside (out) and inside (in) of the cell membrane are indicated and are the same in (B–F). Evidence is based mostly on coimmunoprecipitation of solubilized proteins ex vivo (see text and Table 1 for more details). In each case, presumed sites of protein-protein interaction are shown with pink ovals. (A) The extracellular loops of the voltage-gated calcium channel α₁ subunit (blue) bind to the VWA domain of α₂δ-1 (Gurnett et al., 1997; Wu et al., 2016). Note that this appears to require a GPI-linked α₂δ-1 protein (Davies et al., 2010). (B) The EGF-like (epidermal growth factor) repeats of the thrombospondin trimer (thrombospondin-1 and -2, light green) or pentamer (thrombospondin-4, data not shown) bind to VWA to signal across the membrane via the α₂δ-1 transmembrane domain (yellow) to intracellular proteins, including Rac1 (red arrow), requiring a transmembrane variant of α₂δ-1 (Risher et al., 2018). (C) An unknown extracellular part of α₂δ-1 interacts with the soluble ectodomain regions LG-1 and LG-5 (laminin-like globular domains) of neurexin-1α (Tong et al., 2017). This interaction reduces radioligand binding of gabapentin to α₂δ-1 (Martínez San Segundo et al., 2020), and gabapentin applied to this complex signals to the presynaptic cytosol to reduce the effective size of the readily releasable pool of vesicles (Martínez San Segundo et al., 2020). (D) The α₂δ-1 transmembrane domain interacts with an unknown sequence on NMDA receptor NR1/NR2A and NR1/NR2B proteins (red) (Chen et al., 2018) to alter NMDA receptor function. (E) BK-type calcium-dependent potassium channels (bright green) associate with α₂δ-1 in a mutually exclusive manner with calcium channel α₁ (Zhang et al., 2018), but effects of gabapentinoid drugs on BK channels (if any) are not known. (F) LRP1 (blue) binds to the α₂δ-1 VWA domain via two extracellular ligand binding domains (Kadurin et al., 2017).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Representative electron micrographs ipsilateral (C and D) and contralateral (E and F) to spinal nerve damage. Dots show immunogold labeling of α₂δ-1 in the dorsal horn (lamina I–III) of L4 and L5 spinal cord sections 14 days after sensory nerve ligation that causes allodynia. Scale bar, 0.5 μm. (Arrowheads show presynaptic sites at the plasma membrane of excitatory axon terminal boutons (“b”), characterized by round synaptic vesicles and postsynaptic density). Arrows show postsynaptic sites at the extrasynaptic plasma membrane of dendritic shafts (Den); double arrows show intracellular dendritic sites. (G) Ratio of presynaptic to postsynaptic numbers of particles on the sides ipsilateral (filled bars) and contralateral (open bars) to injury. Figure reproduced from Bauer et al. (2009) with permission (copyright 2009, Society for Neuroscience).

NMDA Receptor Interaction with α₂δ-1

Early in the investigation of the mechanism of action of gabapentin, it was suggested that NMDA receptors in brain and spinal cord were mediators of its pharmacology (Oles et al., 1990; Singh et al., 1996; Jun and Yaksh, 1998; Yoon and Yaksh, 1999), but this suggestion was generally overlooked. Results with gabapentinoid drugs in models of NMDA receptor function are summarized in Table 3. The early studies showed that the analgesic, anticonvulsant, and anxiolytic actions of gabapentin in rodent models were prevented by prior intracerebroventricular injection with the NMDA receptor glycine site agonist d-serine but not its inactive stereoisomer l-serine (d-serine injections alone did not alter behavioral responses). These findings with animal models resembled those with HA-966 (Singh et al., 1990), an experimental drug with partial agonist actions at the NMDA receptor glycine site. In addition, an unpublished doctoral thesis showed that isolated rat striatal neurons have electrical responses to NMDA plus glycine that were reduced by gabapentin, and gabapentin alone behaved like a glycine site partial agonist (Sprosen, 1991). However, these authors noted that gabapentin did not displace radiolabeled glycine binding at NMDA receptors, and d-serine did not interact with gabapentin radioligand binding at α₂δ-1, so they concluded that the gabapentin-NMDA receptor interaction, if real, must be indirect (Singh et al., 1996). Pregabalin also was later shown not to interact with binding sites on NMDA receptors (Li et al., 2011). Once it was shown that drug binding to α₂δ-1 was required for the pharmacology of gabapentin and pregabalin, the early findings regarding gabapentin and NMDA receptor responses were mostly forgotten. However, it is now clear that, in addition to calcium channels, α₂δ-1 interacts with NMDA receptors and several other proteins.

NMDA Receptor Proteins Bind Directly to α₂δ-1

NMDA receptors consist of a tetramer of both NR1 and NR2 subunits (NR1/NR2A and NR1/NR2B are most common in brain, with occasional NR1/NR3), and they are located both presynaptically and postsynaptically in spinal cord and brain (Traynelis et al., 2010). Chen and coworkers from the H.L. Pan laboratory (Chen et al., 2018) showed with coimmunoprecipitation that α₂δ-1 interacts with the NMDA receptors of GluN1/2A and GluN1/2B types by binding that requires a specific sequence on the GluN1 subunit. This physical interaction between proteins also requires complete multimeric NMDA receptor proteins because single subunits (GluN1, GluN2A, or GluN2B) did not interact with α₂δ-1. The interaction occurs between the membrane-spanning C-terminal region of α₂δ-1 and an unidentified region of NMDA receptor heteromers that includes the NR1 subunit. These interactions between α₂δ-1 and NMDA receptors do not require the widely studied von Willebrand domain of the α₂δ-1 protein, in contrast to the interaction between α₂δ-1 and thrombospondin (Eroglu et al., 2009) or between α₂δ-1 and calcium channel α₁ subunits (Canti et al., 2005). Furthermore, studying fluorescent resonance energy transfer (FRET) between tagged α₂δ-1 proteins and tagged NMDA receptors in model cells, Chen et al. (2008) showed that FRET (which is absent unless the two proteins are physically adjacent) is absent in the presence of gabapentin (Chen et al., 2018). Additional data suggest that gabapentin decreases the physical interaction between α₂δ-1 and NMDA receptors in this system and also disrupts traffic of both α₂δ-1 and NMDA receptors from the cytosol to the cell membrane. In addition, experimental coexpression of α₂δ-1 with GluN1/2A alters NMDA receptor properties such that much more inward current occurs at membrane potentials between −80 and −20 mV; that is, α₂δ-1 reduces magnesium block of the NMDA receptor channel, which ordinarily prevents current from flowing at resting potentials. Magnesium block ordinarily prevents NMDA receptors from functioning at resting membrane potentials and causes NMDA receptors to function as coincidence detectors, only active with simultaneous glutamate activation and cell membrane depolarization (Traynelis et al., 2010). The reduced magnesium block was not seen when α₂δ-1 was coexpressed with GluN1/2B receptors but was only seen with GluN1/2A receptors (Chen et al., 2018). Importantly, the effect of α₂δ-1 on magnesium block was completely absent in the presence of gabapentin. These results suggest that association of NMDA GluN1/2A receptors with α₂δ-1 would cause more than the normal amount of glutamate-gated current near resting membrane voltages, but this augmentation of NMDA receptor responses would be reduced by gabapentinoid drugs.

For cases in which it was tested, the ability of gabapentin and pregabalin to reduce spontaneous vesicle release in chronic-pain models was absent if NMDA receptors were blocked in spinal cord dorsal horn neurons (Chen et al., 2018, 2019; Deng et al., 2019). Similarly, pregabalin reduced spontaneous vesicle release (measured by release of fluorescent dye from synaptic vesicles) from synapses in cultured hippocampal neurons (Micheva et al., 2006), but not with NMDA receptors blocked. Since α₂δ-1 is expressed mostly presynaptically, it is likely that presynaptic membranes are an important anatomic site of α₂δ-1 interactions with NMDA receptors, as suggested experimentally (Chen et al., 2018, 2019; Zhou et al., 2018; Deng et al., 2019). However, it will require additional studies to conclusively show that presynaptic NMDA receptors are uniquely sensitive to gabapentinoids. In summary, it is likely that gabapentin and pregabalin reduce spontaneous (miniature) synaptic release in several regions of brain, a manner that requires α₂δ-1-linked presynaptic NMDA receptors.

Although considerable evidence links gabapentinoid drugs to some NMDA receptors, the anatomic distribution of immunostained NMDA receptors in brain (Petralia et al., 1994) is quite different from that for immunostained α₂δ-1 (Taylor and Garrido, 2008). For example, NMDA receptor immunostaining is most dense in cell body layers of hippocampus (where α₂δ-1 immunostaining is sparse), with denser α₂δ-1 immunostaining in dendritic layers. Furthermore, a survey of gabapentinoid actions on NMDA receptor–dependent processes (Table 3) shows both inhibitory effects and negligible effects in different preparations. Therefore, it is clear that α₂δ-1 proteins and gabapentinoid drugs interact only with a subset of NMDA receptors.

Analgesia Produced by Gabapentinoids Compared with Known NMDA Antagonists

NMDA antagonists acting at all NMDA receptors have pharmacological profiles very different from the gabapentinoids and are far from ideal analgesic, antiseizure, or anxiolytic drugs. To date, no broad-spectrum NMDA antagonist drug has proven useful for treating epilepsy or chronic pain in humans, likely because broad-spectrum NMDA antagonists have undesirable properties consisting of feeling dissociated from present time and surroundings (dissociative effects) (van Schalkwyk et al., 2018), memory disruption, confusion, agitation, nausea, and sometimes psychosis, particularly at high dosages. This contrasts with gabapentin and pregabalin, which do not notably cause dissociative effects, disrupt memory, or cause agitation or psychosis in clinical use. Although ketamine (the most widely studied NMDA antagonist for treating pain) is used primarily for perisurgical pain (Kreutzwiser and Tawfic, 2019) and increasingly for treatment-resistant depression (Murrough et al., 2013; van Schalkwyk et al., 2018), it is not used for chronic neuropathic pain (Gilron, 2007), and it has strong psychotomimetic properties (Sos et al., 2013).

In contrast to recent findings that pregabalin reduces NMDA receptor responses in some regions (e.g., Zhou et al., 2018), studies show that gabapentin and pregabalin have variable effects on long-term potentiation (LTP) of glutamate synapses (Table 3), with most studies showing either no effect or a very modest action of these drugs on the formation of LTP, which is widely accepted to require activation of postsynaptic NMDA receptors.

Despite these findings, the H.L. Pan laboratory recently showed that, in vitro, gabapentin completely prevents NMDA receptor–dependent LTP of neocortical afferents to the dorsal striatum (Zhou et al., 2018) by an action on α₂δ-1 that modulates both presynaptic and postsynaptic NMDA receptors. It was previously shown that corticostriatal LTP requires presynaptic NMDA receptors (Park et al., 2014), and it has become clear that presynaptic NMDA receptors have different properties from the more widely studied postsynaptic NMDA receptors of neocortical and hippocampal neurons (Banerjee et al., 2016; Dore et al., 2017; Bouvier et al., 2018). Other findings suggest that gabapentin may reduce presynaptic NMDA receptor activity in area CA1 of rat hippocampal slices (Suarez et al., 2005) and in entorhinal cortex slices, where gabapentin reduced the occurrence of spontaneous miniature synaptic events (Cunningham et al., 2004). Furthermore, both gabapentin and pregabalin inhibit cortical spreading depression in vitro (Cain et al., 2017), which might result from inhibiting presynaptic NMDA receptors (Zhou et al., 2013). In summary, there is solid evidence of an interaction between α₂δ-1 drugs and NMDA receptor function, but it appears that inhibition is restricted to only a few brain locations and/or at a molecular subset of NMDA receptors (Chen et al., 2018; Luo et al., 2018; Ma et al., 2018; Zhou et al., 2018).

Some recent authors (Chen et al., 2018) have proposed that an α₂δ-1-NMDA receptor interaction is required for analgesic actions of gabapentin and pregabalin in animal models and in clinical use. However, this has only been demonstrated in the sensory nerve ligation model of neuropathic pain with spinal reflex-like tactile responding in mice. Analgesia with gabapentin or pregabalin for pain-related responses in animals that are not reflex-like and that require the forebrain have not been studied to see if they rely upon α₂δ-1 interactions with NMDA receptors. See, for example, Bannister et al., (2017).

Furthermore, it is clear that NMDA receptors are not required for all of the actions of gabapentin or pregabalin. For example, gabapentin rapidly reduces the frequency of spontaneous miniature excitatory synaptic potentials in dorsal horn neurons of sensory nerve ligated mice (an apparently presynaptic action), even in the presence of an NMDA antagonist (Zhou and Luo, 2015). Nevertheless, it is quite interesting that gabapentin alters NMDA receptor function in animal models of acute focal ischemia (Luo et al., 2018) and at corticostriate glutamate afferent inputs to the striatum (Zhou et al., 2018), including LTP of the corticostriate pathway.

In summary, NMDA receptors newly incorporated into neuronal membranes after ischemia (Luo et al., 2018), neuropathic pain stimulation (Chen et al., 2018, 2019), or spreading depression (Cain et al., 2017) may be particularly sensitive to gabapentin and pregabalin, as are certain native NMDA receptors located in brain regions such as the striatum (Zhou et al., 2018).

α₂δ-1 Proteins Interact with Only Certain NMDA Receptors

Presynaptic NMDA receptors (Banerjee et al., 2016; Abrahamsson et al., 2017; Bouvier et al., 2018) have been studied between nearby pyramidal neurons of neocortex, on corticoamygdalar glutamatergic endings, and in long-term depression in the hippocampus that is timing-dependent. These receptors are important for use-dependent facilitation of glutamate release (Woodhall et al., 2001; Li et al., 2008, 2009), and presynaptic NMDA receptors have specific protein subunits that differ from those of the more widely studied postsynaptic NMDA receptors. For example, mature mice have predominantly GluN1/GluN2B subunits at presynaptic NMDA receptors between neocortical and hippocampal neurons (Woodhall et al., 2001; Larsen et al., 2011), whereas postsynaptic GluN1/GluN2A and GluN1/GluN2B receptors both contribute to LTP in hippocampus (Liu et al., 2004; Berberich et al., 2005). Glutamate receptors consisting of GluN1/GluN2B subunits are also required for long-term depression in hippocampus (Liu et al., 2004). Although details are not yet very clear, it is likely that gabapentinoids and NMDA antagonists differ functionally by acting at distinct subpopulations of NMDA receptors. Previous studies suggested that some cellular actions of gabapentin and pregabalin require activation of protein kinases (Gu and Huang, 2001; Maneuf and McKnight, 2001; Fehrenbacher et al., 2003), and it is clear that enhanced NMDA receptor function from phosphorylation contributes to neuropathic and chronic pain (Salter and Kalia, 2004; Salter and Pitcher, 2012). It has not yet been studied whether alternative splicing (Sengar et al., 2019) of NMDA receptors alters the interaction with α₂δ-1 and modulation by gabapentinoid drugs. However, a recent paper (Huang et al., 2020) indicates that increased NMDA receptor phosphorylation in spinal cord enhances the amount of α₂δ-1 bound NMDA receptor protein and also increases the gabapentin sensitivity of NMDA receptors.

Upregulation of α₂δ-1 and NMDA Receptor Function After Neuropathic Injury

It has been known for 20 years that spinal α₂δ-1 protein is markedly upregulated after peripheral nerve injury and that this upregulation causes neuropathic pain symptoms in animal models (Luo et al., 2001; Boroujerdi et al., 2011; Gong et al., 2018). More recently, electron microscopy has revealed that the change in α₂δ-1 density occurs primarily in presynaptic sensory neurons rather than postsynaptic dendrites (Bauer et al., 2009) (Fig. 3). That change is accompanied by increased spinal neuron responses to application of NMDA and also by increases in the frequency of glutamate-dependent miniature synaptic potentials in dorsal horn sensory neurons (Chen et al., 2018). Both the increased NMDA receptor–mediated responses and increased NMDA receptor–mediated miniature synaptic potentials are blocked by acute application of gabapentin or pregabalin. Therefore, in at least some models of neuropathic pain, there is both upregulation of α₂δ-1 protein and upregulation of NMDA receptor function in the spinal cord.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Proposed gabapentin-sensitive interactions of α₂δ-1 with synaptic proteins other than calcium channels. (A) Gabapentin interferes with α-neurexin at synapses in model cells to reduce the size of the readily releasable pool (RRP) of presynaptic vesicles (Martínez San Segundo et al., 2020). (B) Gabapentin interferes with the action of presynaptic NMDA receptors in several systems (Chen et al., 2018, 2019; Luo et al., 2018; Ma et al., 2018; Zhou et al., 2018; Deng et al., 2019) to reduce the spontaneous release of synaptic vesicles and decrease NMDA receptor function. (C) Gabapentin interferes with the action of thrombospondin from astrocytes, and this reduces presynaptic vesicle release and also reduces postsynaptic spine enlargement (Risher et al., 2018; Wang et al., 2020). The effect of gabapentin to prevent spine enlargement is mediated by the small Rho GTPase protein, Rac-1, and requires activation of Rac-1 by guanine exchange factors and also requires NMDA receptors. This process then activates the actin cytoskeleton.

Neurexin-1α as an Additional Target of Gabapentinoids

Neurexins are a family of presynaptic proteins that protrude from presynaptic terminals into the extracellular space and that have diverse functions in different neuron types. The neurexins are expressed both as full-length membrane-spanning protein (α-neurexin) and truncated (β-neurexin) versions. Neurexins form an important part of the trans-synaptic network that modulates synapse functions (Sudhof, 2017). Neurexins prominently bind to postsynaptic proteins of the neuroligin family (Miller et al., 2011) to stabilize and spatially align pre-and postsynaptic elements. Both proteins interact with a large number of other synaptic scaffolds, postsynaptic receptors, and presynaptic proteins (Biederer et al., 2017; Sudhof, 2017).

Cell cultures of neurons lacking native neurexins have reduced synaptic calcium influx and transmitter release that is restored by α-neurexin expression. Surprisingly, this effect of α-neurexin to enhance synaptic function appears to be mediated by a weak interaction between neurexin and α₂δ-1, at least in one model system (Brockhaus et al., 2018). Interestingly, these effects of α₂δ-1 on neurexin function did not appear to be mediated by a stable protein-protein interaction but, rather, by a somewhat weak and transient interaction (Brockhaus et al., 2018).

More recently, it was shown that the soluble extracellular domain of neurexin-1α reduces radioligand binding of [³H]gabapentin to recombinant α₂δ-1, indicating a direct interaction between the two proteins (Martínez San Segundo et al., 2020). Functional studies of this interaction were done with a model synapse system cultured in vitro. Microcultures of individual rat superior cervical autonomic neurons form monosynaptic acetylcholine synapses back onto themselves when grown in this manner. These “autaptic” synapses were studied using electrophysiology and presynaptic calcium imaging of synaptic boutons. Application of pregabalin (30 μM) reproducibly and reversibly reduced neurotransmitter release by about 50%, as measured by postsynaptic current. The pregabalin effect was particularly pronounced in response to rapid trains of presynaptic action potentials, and these measurements indicated that pregabalin reduced the size of the “readily releasable pool” of synaptic vesicles (Rosenmund and Stevens, 1996; Kaeser and Regehr, 2017). Importantly, the effect of pregabalin was entirely independent of changes in the function of presynaptic voltage-gated calcium channels, since presynaptic calcium influx (measured by a fluorescent indicator) was unchanged. Finally, it was shown that extracellular application of a soluble fragment of neurexin-1α reduced neurotransmitter release to the same extent as pregabalin and occluded the effects of pregabalin, suggesting that pregabalin and neurexin alter synapse function through the same molecular pathway.

This investigation (Martínez San Segundo et al., 2020) was aided by several technical advantages over other studies of gabapentinoids. The hardiness of autonomic ganglion neurons allowed long-term recordings of synaptic currents without much postsynaptic receptor plasticity (a complicating factor at hippocampal synapses, for example). This allowed estimation of the time course of drug effects (onset time of about 5 minutes and washout time constant of about 50 minutes) and straightforward estimation of readily releasable pool size. The use of a genetically coded presynaptic calcium indicator was possible since only synapses from a single cell were present in each microculture. These results show that, at least in this model system, pregabalin reduces synaptic strength by reducing the readily releasable pool via an interaction between α₂δ-1 and presynaptic α-neurexin, without any change in the function of presynaptic calcium channels. It will require additional work to establish whether the changes mediated by neurexin are relevant for analgesia by gabapentinoid drugs.

Thrombospondins as an Additional Target of Gabapentinoids

Thrombospondins are a family of extracellular matrix proteins [reviewed in Risher and Eroglu (2012)] that in the brain are formed mostly by astrocytes. Astrocytes, both in spinal cord and neocortex, are thought to play a major role in the pathogenesis of chronic pain (Hansen and Malcangio, 2013). Thrombospondins are released from astroctyes into the extracellular space in response to several stimuli, including activation of glial P2X purinergic gated ion channel receptors (Tran and Neary, 2006; Kim et al., 2016) that respond to ATP from damaged cells or are released from nearby glial calcium waves (Guthrie et al., 1999; Haydon and Carmignoto, 2006). The P2X purinergic gated ion channel receptor family play a significant role in chronic pain (Chizh and Illes, 2001). The α₂δ-1 protein was shown to interact with thrombospondin (particularly subtypes 1 and 4) by immunoprecipitation (Eroglu et al., 2009) and by interaction between purified proteins (Park et al., 2016, 2018) (see Table 1). However, two separate studies have failed to show that gabapentin directly disrupts the molecular interaction between thrombospondin and α₂δ-1 proteins in vitro (Lana et al., 2016; El-Awaad et al., 2019), suggesting that the inhibition by gabapentin of thrombospondin actions could be indirect or require the presence of other proteins. Additional studies suggest that signaling proteins, including the scaffolding protein LRP1 (Kadurin et al., 2017) or activation of the small Rho GTPase, Ras-related C3 botulinum toxin substrate 1 (Rac1), are part of the biochemical pathway involved in the synaptogenic action that is inhibited by gabapentinoid drugs (Risher et al., 2018). Interestingly, LRP1 also was found to interact with β1-integrins at the cell surface to regulate their function (Theret et al., 2017).

Gabapentin and pregabalin inhibit several thrombospondin effects via α₂δ-1 and prevent the formation of new synapses in response to thrombospondin application or astroctye activation. It has been known for some time that astrocyte-conditioned media promote the formation of new glutamate synapses in cultured neurons, and it was found that this occurs because α₂δ-1 proteins act as neuronal thrombospondin receptors (Eroglu et al., 2009). After neuropathic sensory nerve injury (Kim et al., 2012) or spinal cord injury (Zheng et al., 2005), thrombospondin induces chronic-pain–like states and increases the number of excitatory synapses and the rate of glutamatergic mEPSCs in spinal dorsal horn in animal models (Nguyen et al., 2009; Kim et al., 2012).

The newly formed synapses in response to thrombospondin are initially silent glutamate synapses (with NMDA receptors but no functional α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid type glutamate receptors) (Eroglu et al., 2009; Risher et al., 2018; Yu et al., 2018; Wang et al., 2020). Recent findings indicate that pain-related responses in mice from overexpression of either α₂δ-1 or thrombospondin-4 can be acutely blocked with gabapentin or pregabalin (Park et al., 2016) and also with a peptide fragment that prevents the interaction between α₂δ-1 and thrombospondin; this peptide also blocks analgesia with pregabalin in a spinal nerve injury model in vivo (Park et al., 2018). Although repeated treatment with gabapentinoids reduces the formation of new synapses induced by thrombospondin, once new synapses formed, gabapentin applied subsequently did not reduce the number or size of synapses but did effectively reduce pain-related behaviors. Therefore, there appear to be two distinct functions of drugs acting at α₂δ-1/thrombospondin: one long-lasting, involving synaptic size, and the other more rapid and reversible, acutely reducing pain-related behaviors.

The effects of thrombospondin to promote chronic-pain–like states are prevented by repeated doses of gabapentin acting at α₂δ-1 either in the spinal cord (Crosby et al., 2015; Pan et al., 2015; Park et al., 2016) or in the somatosensory cortex (Kim et al., 2016). A separate study showed that thrombospondin-4 application to isolated sensory (dorsal root ganglion) neurons for 4 hours alters voltage-gated calcium currents recorded at the cell body (Pan et al., 2016a). Although gabapentin itself did not acutely reduce voltage-gated calcium currents in these cells, it did prevent the long-term effects of thrombospondin-4 to decrease high-voltage activated currents and increase low-voltage activated currents. These results strongly suggest that some actions of gabapentinoid drugs to reduce chronic pain may be independent of reducing calcium channel function.

The idea that a thrombospondin-α₂δ-1 interaction might underlie gabapentinoid drug actions other than analgesia in chronic pain has also been studied. Repeated prophylactic gabapentin treatment in models of post-traumatic epilepsy (Li et al., 2012; Andresen et al., 2014; Lau et al., 2017; Takahashi et al., 2018) prevented the stabilization of abnormal new synapses in neocortex, presumably by blocking the action of astrocyte-derived thrombospondin.

A recent study in the striatum of mice showed that selective stimulation of astrocytes caused the formation of new and stronger glutamate synapses onto striatal medium spiny neurons. This synaptogenesis was mediated by glial-derived thrombospondin-1 acting at α₂δ-1, whose effects were blocked by gabapentin (Nagai et al., 2019). The formation of stronger synapses in striatum was associated with behavioral hyperactivity and disrupted attention, and these abnormal behaviors were also prevented by repeated prophylactic gabapentin treatment. An additional study in mice shows that in the nucleus accumbens shell (an area implicated in drug addiction), cocaine administration triggers the formation of new silent glutamate synapses via thrombospondin-2, and this process is blocked by repeated gabapentin administration acting at α₂δ-1 (Wang et al., 2020). Finally, an important study (Risher et al., 2018) shows that in mouse neocortex, repeated gabapentin treatment reduces the normal formation of new corticocortical synapses by thrombospondin, and this occurs by a postsynaptic interaction between thrombospondin and α₂δ-1, which requires both NMDA receptors and the GTPase Rac1. In this study, activation of the thrombospondin-α₂δ-1 pathway had no effect on GABA synapses. Finally, a recent study (https://ssrn.com/abstract=3470401) shows that prophylactic treatment with gabapentin in an animal model of spinal cord injury prevents plasticity in spinal autonomic circuits that cause autonomic dysreflexia by preventing thrombospondin-induced changes.

In summary, α₂δ drugs may prevent the stabilization of new or stronger glutamate synapses formed specifically in response to astrocyte activation and the subsequent release of thrombospondin. This astrocyte-activated process of synaptic strengthening may occur in response to neuronal damage in neocortex (from release of glutamate, potassium ions, and ATP), by sustained astrocyte activation via GABA_B receptors in striatum, by neuronal activation by cocaine in nucleus accumbens, and also by normal activation of glia during synaptogenesis. Each of these processes appears to activate astrocytes and strengthen nearby glutamate synapses in a process that is blocked by gabapentinoid drugs. It seems likely that the interaction between gabapentinoid drugs and thrombospondin function may be relevant for preventing long-lasting pain-related anatomic changes. It is less clear that the interaction between α₂δ-1 and thrombospondin is necessary for short-lasting analgesia from acute treatment with gabapentinoid drugs in some animal models or clinical use.

Other α₂δ-1 Binding Proteins

BK-type voltage-gated potassium channels compete with voltage-gated calcium channel α1 subunits for α₂δ-1 protein binding (Zhang et al., 2018). BK channels directly associate with Ca_V2.1 and Ca_V2.2 channels at presynaptic endings in brain (Berkefeld et al., 2006; Dai et al., 2009), putting them within a few nanometers of vesicle release machinery at synapses. This localization is required for BK channel function to rapidly hyperpolarize cells in response to presynaptic calcium influx and to moderate calcium-induced vesicle release. Although not studied by protein-protein interaction techniques, recent findings (Hoppa et al., 2014) indicate that changes in α₂δ-1 expression in cultured neurons alter the function of presynaptic K_V1 and K_V3.1 potassium channels. To date, no studies of gabapentin or pregabalin have been published on BK channel function or K_V1/K_V3.1 channel function at synaptic endings.

Brain α₂δ-1 also interacts directly with the membrane-bound protein trafficking molecule LRP1, a transmembrane protein that is involved in mediating α₂δ-1 effects on calcium channel traffic to and from the membrane (Kadurin et al., 2017). LRP1 has many protein-interacting domains and is known to interact with a variety of other proteins (Lillis et al., 2008), including the extracellular matrix proteins fibronectin and thrombospondin; apolipoprotein E; and intracellular proteins such as RAP, Shc (Src homology and collagen family of adaptor proteins), protein kinase C, PSD-95 (post-synaptic density protein of 95 kD MW, which plays an important role with postsynaptic NMDA receptors), and the endoplasmic reticulum protein calreticulin. Calreticulin is involved in protein processing in endosomes. It is possible that LRP1 participates in the interaction between α₂δ-1 and thrombospondin or other proteins like the cell signaling protein RAC1 (Risher et al., 2018).

There is evidence that α₂δ-1 interacts with several other proteins in brain (Table 2), although many of these interactions are very likely indirect, particularly those involving presynaptic vesicle release proteins that are known to directly interact with calcium channel α₁ subunits [reviewed in He et al. (2018)].

Summary and Conclusions

Since the discovery of the analgesic activity of gabapentin, identifying its mechanism of action at the cellular level has been challenging, with many promising and plausible, but apparently false, leads. It is now clear that, although the analgesic effects of gabapentin-like drugs involve an interaction with α₂δ-1, those effects clearly are not limited to voltage-gated calcium channels. Instead, there is compelling new evidence that several gabapentinoid drug effects involve other proteins that interact with α₂δ-1, specifically a subset of NMDA receptors, neurexin-1α, and thrombospondins (Fig. 4).

Interactions between α₂δ-1 and neurexins could explain at least part of the gabapentinoid drug’s ability to reduce neurotransmitter release from neocortical and hippocampal tissues, where there is not always a clear correlation between neurotransmitter release and decreased calcium channel function. In light of the findings reviewed here about interactions of α₂δ-1 with multiple synaptic proteins and findings that overexpression of α₂δ-1 in cultured neocortical neurons actually decrease presynaptic calcium influx (Hoppa et al., 2012, 2014), it seems likely that synaptic proteins other than calcium channel α1 subunits may be required for gabapentinoid drugs to reduce mEPSC frequency and excitatory neurotransmitter release.

The interaction between α₂δ-1 and NMDA receptors is particularly intriguing because NMDA receptors have long been seen as promising targets for analgesic and antiseizure drugs but have proven elusive in terms of drugs with favorable risk/benefit profiles. Although gabapentin and pregabalin clearly do not modulate all NMDA receptors, it is also clear that some subsets of NMDA receptors are modulated by drug binding to α₂δ-1. The interaction of thrombospondins with α₂δ-1 may also be important for some pharmacological actions of gabapentin and pregabalin, particularly when pain is augmented by activation of glia that release thrombospondin to form new and enlarged glutamate synapses.

An important unanswered question with gabapentinoid drugs is whether different proteins can interact at α₂δ-1 at the same time or whether such interactions are mutually exclusive. In this regard, one study has shown that BK potassium channels compete with calcium channel α₁ subunits for binding α₂δ-1. However, it is not yet known whether α₂δ-1 interactions with thrombospondins, NMDA receptors, or neurexins are mutually exclusive with each other or with calcium channel α₁ proteins. An important step in this direction was provided by findings that a GPI anchor at the membrane portion of α₂δ-1 is required for increasing cellular calcium currents (Davies et al., 2010), but a similar modification prevents the α₂δ-1/thrombospondin action that enlarges synaptic spines (Risher et al., 2018). Furthermore, a full-length and apparently membrane-spanning α₂δ-1 is required for α₂δ-1-NMDA receptor interactions (Chen et al., 2018). Thus, differently processed α₂δ-1 proteins are required for some of the different gabapentinoid drug functions in neurons.

In conclusion, although gabapentinoids do appear to act by binding to α₂δ-1, analgesic effects appear to involve not only voltage-gated calcium channels but also other proteins, including a subset of NMDA receptors, neurexins, thrombospondins, and possibly other proteins. In particular, existing evidence suggests the α₂δ-1-thrombospondin interaction independent of calcium channels may be important for gabapentinoid drugs to reduce long-lasting pain with only acute treatment around the time of nerve injury.