Aberrant glutamatergic systems underlying impulsive behaviors: insights from clinical and preclinical research

Abstract

Impulsivity is a broad construct that often refers to one of several distinct behaviors and can be measured with self-report questionnaires and behavioral paradigms. Several psychiatric conditions are characterized by one or more forms of impulsive behavior, most notably the impulsive/hyperactive subtype of attention-deficit/hyperactivity disorder (ADHD), mood disorders, and substance use disorders. Monoaminergic neurotransmitters are known to mediate impulsive behaviors and are implicated in various psychiatric conditions. However, growing evidence suggests that glutamate, the major excitatory neurotransmitter of the mammalian brain, regulates important functions that become dysregulated in conditions like ADHD. The purpose of the current review is to discuss clinical and preclinical evidence linking glutamate to separate aspects of impulsivity, specifically motor impulsivity, impulsive choice, and affective impulsivity. Hyperactive glutamatergic activity in the corticostriatal and the cerebro-cerebellar pathways are major determinants of motor impulsivity. Conversely, hypoactive glutamatergic activity in frontal cortical areas and hippocampus and hyperactive glutamatergic activity in anterior cingulate cortex and nucleus accumbens mediate impulsive choice. Affective impulsivity is controlled by similar glutamatergic dysfunction observed for motor impulsivity, except a hyperactive limbic system is also involved. Loss of glutamate homeostasis in prefrontal and nucleus accumbens may contribute to motor impulsivity/affective impulsivity and impulsive choice, respectively. These results are important as they can lead to novel treatments for those with a condition characterized by increased impulsivity that are resistant to conventional treatments.

1. Introduction

Impulsivity is a broad construct that encompasses multiple behaviors and can be assessed with self-report questionnaires such as the Barratt Impulsiveness Scale (BIS-11) (Patton et al., 1995) and the UPPS Scale (Whiteside and Lynam, 2001) and with cognitive-behavioral tasks (these tasks will be described in Section 1.1.). Impulsive behaviors are problematic in certain contexts and are defining features of psychiatric conditions such as attention-deficit/hyperactivity disorder (ADHD) and impulse-control disorders like oppositional defiant disorder, conduct disorder, and intermittent explosive disorder (American Psychiatric Association, 2013). In addition to ADHD and impulse-control disorders, impulsive behaviors are observed in psychiatric conditions like schizophrenia, mood disorders, personality disorders (e.g., borderline personality disorder), substance use disorders, and behavioral addictions (Amlung et al., 2019; Evenden, 1999; Hoptman, 2015; Jentsch et al., 2014; Lee et al., 2019). As a multifaceted construct, the current review will focus on three major forms of impulsivity: motor impulsivity, impulsive choice, and affective impulsivity.

1.1. Major Types of Impulsivity

Motor impulsivity (i.e., impulsive action) is often measured with tasks that require individuals to engage in either action cancellation or action restraint. In stop signal reaction time tasks (SSRTTs), participants are required to inhibit responses they have already initiated when presented with a cue (Logan and Cowan, 1984; Logan et al., 1984; Verbruggen and McLaren, 2018). In go/no-go tasks, individuals are required to either initiate a response (go) or inhibit a response (no-go) when presented with different cues (Newman et al., 1985). In a differential reinforcement of low rates of responding (DRL) task, responses are reinforced only if a response occurs after a designated amount of time. For example, when a DRL-10 s schedule is used, a response is reinforced if the subject waits at least 10 s before emitting a response. A premature response causes the timer to reset (Ferster and Skinner, 1957).

Impulsive choice is primarily conceptualized as the inability to delay gratification; in other words, individuals may forgo a better outcome that is delayed in favor of an immediate, but less preferable, alternative. Delay-discounting tasks are commonly used to measure impulsive choice (Rachlin et al., 1991). Individuals are asked to indicate their preference for a small amount of money made available immediately or a large amount of money delivered after a delay (e.g., $100 now vs. $1000 in 2 months). Steeper discounting indicates higher impulsive choice as individuals are unwilling to wait to earn larger monetary rewards.

Affective impulsivity relates to impulsive behaviors that are dictated by strong emotional responses. One form of affective impulsivity is urgency, in which an individual acts rashly while experiencing a highly negative or a highly positive emotion (Cyders and Smith, 2008). Reactive aggression results from a situation in which the individual feels they have little control; the goal is to neutralize the aversive stimulus (e.g., voluntary manslaughter, lashing out at a supervisor following a negative performance review, etc.). This form of aggression contrasts with proactive aggression, which is premediated, with the aim of achieving an internal or an external goal (e.g., murder, armed robbery, sexual assault, etc.) (see Wrangham, 2018 for a review). Aggression can be directed to oneself in the form of self-injurious behavior, with suicide being a highly concerning form of affective impulsivity (Gvion et al., 2015).

1.2. Animal Models of Impulsivity

To further understand the underlying causes of impulsive behaviors, preclinical research is frequently used. One major area of research centers on the use of animal models of ADHD such as the spontaneously hypertensive rat (SHR) (Sagvolden, 2000) and the Naples High Excitability (NHE) rat (Viggiano et al., 2002). The SHR is an inbred strain derived from the Wistar Kyoto (WKY) rat. Relative to inbred (WKY) and outbred (Wistar, Sprague Dawley) control strains, SHRs show enhanced locomotor activity (Cailhol and Mormède, 1999; Dela Peña et al., 2021; Gungor Aydin and Adiguzel, 2023), impaired attention (Jentsch, 2005; Sagvolden and Xu, 2008), increased impulsive choice (Fox et al., 2008; Fox et al., 2023; Sanabria and Killeen, 2008), and increased motor impulsivity as assessed in a DRL task (Sable et al., 2021). The NHE rat is an outbred strain that has been selectively bred from Sprague Dawley rats since 1976 based on behavioral arousal to novelty. While they display attentional deficits, NHE rats do not display general hyperactivity or impulsive behaviors (Viggiano et al., 2002). As such, discussion of animal models of ADHD will be limited to SHRs in the current review.

Impulsive behaviors can be measured in both SHRs and outbred rodent strains using similar procedures as those used with humans, including the SSRTT (Eagle and Robbins, 2003), go/no-go tasks (Carro-Ciampi and Bignami, 1968; Cole and Michaleski, 1986; Livesey et al., 1980; Sakurai and Sugimoto, 1985), and DRL tasks (Hankosky and Gulley, 2013; Simon et al., 2013; Somkuwar et al., 2016). A popular test of motor impulsivity in rodents is the five-choice serial reaction time task (5CSRTT) (Carli et al., 1983), which is an animal analog of the continuous performance task. While the continuous performance task measures sustained and selective attention, the 5CSRTT measures both attention and motor impulsivity. In the 5CSRTT, rodents must respond in one of several apertures depending on which stimulus light is illuminated. Importantly, rodents must wait a certain amount of time before they respond. Premature responses are used as an index of motor impulsivity. The 5CSRTT can be modified such that the number of active apertures is reduced (1CSRTT, 2CSRTT, etc.; e.g., Abbott et al., 2022; Nord et al., 2019; Tsutsui-Kimura et al., 2009; Ucha et al., 2019).

As with humans, delay-discounting procedures are often used to model impulsive choice (Evenden and Ryan, 1996; Green et al., 2010; Mitchell et al., 2006; Pitts and Febbo, 2004; Rajala et al., 2015; Woolverton et al., 2007). There are several permutations of delay discounting. In most versions of delay discounting, subjects choose between a small, immediate alternative and a large, delayed alternative. The delay to receiving the large magnitude alternative can either increase (Evenden and Ryan, 1996) or decrease (Tanno et al., 2014) within a session or across sessions (Mobini et al., 2002). Alternatively, the delay to receiving the large, delayed alternative can be adjusted depending on prior choice for each alternative (adjusting delay procedures) (Mazur, 1987; Perry et al., 2005). Discussing each version of delay discounting in detail is beyond the scope of this review (see Madden and Johnson, 2010; Yates, 2018 for discussions on the various delay discounting procedures).

1.3. Purpose of Review

As a transdiagnostic trait, there is a large body of literature devoted to better understanding the neurobehavioral mechanisms that control impulsivity. From a neurobiological perspective, considerable attention is directed toward monoaminergic neurotransmitter systems in impulsive behaviors in both humans and animals (Cardinal, 2006; Dalley and Ersche, 2019; Evenden, 1999; Winstanley, 2011). However, there is growing interest in understanding glutamatergic control of several psychiatric disorders, including ADHD (Lesch et al., 2013), schizophrenia (Moghaddam and Javitt, 2012), mood disorders (Jun et al., 2014), and addiction (Kalivas, 2009; Pettorruso et al., 2014). As such, the purpose of the current review is to discuss the contributions of glutamate to impulsive behaviors using clinical and preclinical evidence. Before discussing glutamatergic underpinnings of impulsivity, a brief overview of the glutamatergic system is needed.

2. Overview of Glutamatergic System

Glutamate is the major excitatory neurotransmitter in the mammalian nervous system and is critically involved in long-term potentiation (LTP) and long-term depression (LTD), important cellular mechanisms by which learning occurs (Gladding et al., 2009; Lüscher and Malenka, 2012). Primarily synthesized from glial-released glutamine by the mitochondrial enzyme glutaminase (Torgner and Kvamme, 1990), glutamate is packaged in synaptic vesicles by the vesicular glutamate transporter (Eriksen et al., 2020). Release of glutamate into the synapse is mediated by Ca⁺⁺ influx (Nicholls and Sihra, 1986). Once released into the synapse, glutamate binds to either metabotropic receptors (mGluRs) or ionotropic receptors, with each receptor type being further divided (see Fig. 1 for a schematic of a glutamatergic synapse).

Figure 1.

A schematic of a glutamatergic synapse. The inset depicts the postsynaptic density. Created with BioRender.com.

There are three groups of mGluRs. Group I mGluRs (mGluR₁ and mGluR₅) are near-exclusively located on postsynaptic neurons and are G_q-coupled, stimulating the activation of phospholipase C_β. Group II (mGluR₂ and mGluR₃) and Group III (mGluR₄, mGluR₆, mGluR₇, and mGluR₈) are G_i-coupled, leading to decreased adenylyl cyclase activation. While Group II mGluRs are found on both presynaptic and postsynaptic neurons, Group III mGluRs are predominately isolated to presynaptic terminal buttons (see Niswender and Conn, 2010 for a comprehensive review of mGluRs).

Ionotropic receptors are divided into kainate, AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) and NMDA (N-methyl-D-aspartate) receptors, although kainate receptors are often grouped with AMPA receptors as non-NMDA receptors (e.g., Agrawal and Fehlings, 1997; Scott et al., 2019). Each ionotropic receptor is composed of four subunits surrounding an ion channel. Instead of activating a G protein, stimulation of kainate/AMPA and NMDA receptors allows positively charged ions to pass through the channel and into the neuron (Na⁺ for kainate/AMPA; Ca⁺⁺ and Na⁺ for NMDA). AMPA receptors are composed of combinations of GluA1, GluA2, GluA3, and GluA4 subunits while kainate receptors contain GluR5, GluR6, GluR7, KA1, and KA2 subunits. NMDA receptors are composed of two GluN1 subunits, and they contain combinations of GluN2A, GluN2B, GluN2C, and GluN2D subunits (see Bettler and Mulle, 1995; Gan et al., 2015; Regan et al., 2015 for discussions on ionotropic receptor structure). At resting membrane potential, a Mg²⁺ ion blocks the channel of the NMDA receptor. Depolarization of the neuron via neighboring AMPA receptors causes displacement of the Mg²⁺ ion from the NMDA receptor channel. The NMDA receptor requires two co-agonists: glycine, which binds to the GluN1 subunit, and glutamate, which binds to the GluN2 subunit.

LTP is commonly associated with ionotropic receptors, but mGluRs are also important for this form of synaptic plasticity to occur (Anwyl, 2009). For simplicity, I will focus on NMDA receptor-mediated LTP. In early phase LTP, high-frequency stimulation of glutamatergic neurons leads to Ca⁺⁺ influx into the postsynaptic neuron, which causes phosphorylation of additional AMPA receptors by protein kinases such as calcium/calmodulin-dependent protein kinase II (CaMKII); this, in turn, increases the postsynaptic neuron’s sensitivity to glutamate. In late phase LTP, increased Ca⁺⁺ levels lead to protein synthesis and gene transcription via the activation of other protein kinases like protein kinase A (Bliss and Collingridge, 1993; Madison et al., 1991; Malenka and Bear, 2004). LTD is also mediated by ionotropic glutamate receptors. Low-frequency stimulation of the postsynaptic neuron does not sufficiently displace the Mg⁺⁺ from the channel. Lower levels of Ca⁺⁺ enter the neuron, leading to activation of different signaling pathways that result in downregulation of AMPA receptors (Malenka and Bear, 2004). Low frequency stimulation of postsynaptic mGluRs leads to LTD as well (Lüscher and Huber, 2010).

Ionotropic glutamate receptors and signaling molecules like CAMKII are constituents of the postsynaptic density (PSD), a large collection of proteins located at the postsynaptic membrane of dendrites of excitatory neurons (see inset of Fig. 1). Cell adhesion molecules, such as neural cadherins and neuroligins, are a critical component of the PSD as they contribute to the formation of synapses (Togashi et al., 2009). In addition to cell adhesion molecules, there are multiple scaffolding proteins that are involved in the formation and functional integrity of glutamatergic synapses, including SH3 and multiple ankyrin repeat domains protein (SHANK) (Monteiro and Feng, 2017), Homer proteins (Shiraishi-Yamaguchi and Furuichi, 2007), membrane-associated guanylate kinases (MAGUKs) like PSD-95 (Olsen and Bredt, 2003), and guanylate kinase-associated protein (GKAP) (Kim et al., 1997) (see Kaizuka and Takumi (2018) for a comprehensive discussion of the PSD). At high concentrations, the GTPase activator synaptic Ras GTPase-activating protein 1 (SynGAP) acts as a scaffolding protein (Zeng et al., 2019). During LTP, the PSD is enlarged, and actin protein expression increases, leading to increased dendritic spine growth. This growth allows for the accommodation of increased AMPA receptor expression (Herring and Nicoll, 2016).

Because excessive glutamate levels lead to excitotoxicity (Olney, 1986), extracellular glutamate is tightly regulated. First, membrane-bound transporters, most commonly the excitatory amino acid transporter-2 (EAAT-2) in humans (Divito and Underhill, 2014) and the glutamate transporter-1 (GLT-1) in rodents (Rimmele and Rosenberg, 2016), transfer excess glutamate from the synapse to neighboring astrocytes. Glutamate can then be metabolized by amidation back to glutamine via glutamine synthetase (Schousboe et al., 2014). Second, the cysteine/glutamate antiporter system (x_C⁻) is crucial for maintaining glutamate homeostasis (Warr et al., 1999). The x_C⁻ system exchanges extracellular cystine for intracellular glutamate located in astrocytes; this exchange allows glutamate to bind to presynaptic Group II/III mGluRs, inhibiting the additional release of glutamate from the presynaptic neuron (Bridges et al., 2012).

Given the importance of glutamate to learning and memory, both mGluRs and ionotropic glutamate receptors are widely located in brain regions associated with these processes, including cerebral cortex and hippocampus (Abe et al., 1992; Blackstone et al., 1992; Catania et al., 1994; Monaghan et al., 1984; Tanabe et al., 1992; see Ozawa et al., 1998 for a review). Many glutamatergic neurons originate from the cerebral cortex and send projections to neighboring cortical glutamatergic neurons, as well as to the brainstem, the striatum, and the thalamus, with the latter structure projecting glutamatergic neurons back to the cortex (Schwartz et al., 2012). Glutamatergic pathways will be discussed in more detail in Section 3.5.

3. Glutamatergic Dysfunction Underlying Impulsive Behaviors

There are several methods that can be used to identify glutamatergic dysfunction underlying impulsive behaviors. Genetic association studies are commonly used to determine if certain groups of individuals are more likely to have a specific allele or a haplotype of a gene compared to the general population. See Table 1 for glutamatergic-related genes and their corresponding proteins. Researchers can also quantify glutamate levels using techniques like magnetic resonance spectroscopy (MRS) (common in human studies) or in vivo microdialysis and voltammetry (used in preclinical studies). To increase accuracy in MRS studies, glutamate levels are often measured in conjunction with glutamine (glutamate-glutamine or Glx); additionally, other excitatory neurotransmitters like aspartate can be quantified in addition to glutamate/Glx. Another way to identify glutamatergic aberrations is to compare mRNA or protein expression between individuals that display high impulsivity to those that exhibit low impulsivity. In behavioral pharmacology experiments, subjects are treated with a drug that targets a receptor, a transporter, or the x_C⁻ system. These drugs can be delivered systemically or can be infused into the lateral ventricles or into a specific brain region. Behavioral pharmacology approaches can be used in conjunction with neuroscience techniques, allowing for better determination of how glutamate mediates impulsive behaviors.

Table 1.

Genes related to glutamatergic system in humans

Gene	Chromosomal Location	Corresponding Protein/Enzyme	Function
GRM1	6q24.3	mGluR1	Excitatory metabotropic receptor
GRM2	3p21.2	mGluR2	Inhibitory metabotropic receptor
GRM3	7q21.11-q21.12	mGluR3	Inhibitory metabotropic receptor
GRM4	6p21.31	mGluR4	Inhibitory metabotropic receptor
GRM5	11q14.2-q14.3	mGluR5	Excitatory metabotropic receptor
GRM6	5q35.3	mGluR6	Inhibitory metabotropic receptor
GRM7	3p26.1	mGluR7	Inhibitory metabotropic receptor
GRM8	7q31.33	mGluR8	Inhibitory metabotropic receptor
GRIA1	5q33.2	AMPA subunit GluR1	Subunits of AMPA receptor
GRIA2	4q32.1	AMPA subunit GluR2
GRIA3	Xq25	AMPA subunit GluR3
GRIA4	11q22.3	AMPA subunit GluR4
CACNG2	22q12.3	Calcium voltage-gated channel auxiliary subunit gamma 2	Regulates trafficking and channel gating of AMPA receptors
CACNG3	16p12.1	Calcium voltage-gated channel auxiliary subunit gamma 3
CACNG4	17q24.2	Calcium voltage-gated channel auxiliary subunit gamma 4
CACNG5	17q24.2	Calcium voltage-gated channel auxiliary subunit gamma 5
CACNG7	19q13.42	Calcium voltage-gated channel auxiliary subunit gamma 7
CACNG8	19q13.42	Calcium voltage-gated channel auxiliary subunit gamma 8
GRIN1	9q34.3	NMDA subunit GluN1	Binding site of glycine on NMDA receptor
GRIN2A	16p13.2	NMDA subunit GluN2A	Binding site of glutamate on NMDA receptor
GRIN2B	12p13.1	NMDA subunit GluN2B
GRIN2C	17q25.1	NMDA subunit GluN2C
GRIN2D	19q13.33	NMDA subunit GluN2D
GRIK1	21q21.3	Kainate subunit 1	Subunits of kainate receptor
GRIK2	6q16.3	Kainate subunit 2
GRIK3	1p34.3	Kainate subunit 3
GRIK4	11q23.3	Kainate subunit 4
GRIK5	19q13.2	Kainate subunit 5
SLC17A7	19q13.33	Vesicular glutamate transporter 1	Package intracellular glutamate into vesicles
SLC17A6	11p14.3	Vesicular glutamate transporter 2
SLC17A8	12q23.1	Vesicular glutamate transporter 3
SLC1A3	5p13.2	Excitatory amino acid transporter 1	Reuptake of extracellular glutamate
SLC1A2	11p13	Excitatory amino acid transporter 2
SLC1A1	9p24.2	Excitatory amino acid transporter 3
SLC1A6	19p13.12	Excitatory amino acid transporter 4
SLC1A7	1p32.3	Excitatory amino acid transporter 5
SLC7A11	4q28.3	Cystine-glutamate antiporter	Exchanges cystine for glutamate
GLS	2q32.2	Glutaminase	Synthesizes glutamate from glutamine
GLUL	1q25.3	Glutamine synthetase	Catalyzes synthesis of glutamine from glutamate
GAD1	2q31.1	Glutamic acid decarboxylase	Converts glutamate to gamma-amino butyric acid (GABA)
SHANK1	19q13.33	SHANK1
SHANK2	11q13.3-q13.4	SHANK2
SHANK3	22q13.33	SHANK3
HOMER1	5q14.1	HOMER1
HOMER2	15q25.2	HOMER2
HOMER3	19p13.11	HOMER2
DLG1	3q29	Discs large membrane-associated guanylate kinase (MAGUK) scaffold protein 1	Scaffolding proteins
DLG2	11q14.1	Discs large MAGUK scaffold protein 2
DLG3	Xq13.1	Discs large MAGUK scaffold protein 3
DLG4	17p13.1	Discs large MAGUK scaffold protein 4
DLG5	10q22.3	Discs large MAGUK scaffold protein 5
DLGAP1	18p11.31	DLG associated protein 1
DLGAP2	8p23.3	DLG associated protein 2
DLGAP3	1p34.3	DLG associated protein 3
DLGAP4	20q11.23	DLG associated protein 4
SYNGAP1	6p21.32	Synaptic Ras GTPase activating protein 1

To date, few studies have determined if the glutamatergic system is altered in individuals diagnosed with an impulse-control disorder as defined by the American Psychiatric Association (e.g., oppositional defiant disorder). Although no differences in glutamate or glutamine levels are observed between children and adolescents diagnosed with intermittent explosive disorder relative to controls (Davanzo et al., 2003), more recent evidence suggests that individuals with the C/C genotype of the GRM8 gene are more likely to have conduct disorder compared to individuals with the C/T or the T/T genotypes (Bauer and Covault, 2020). Due to a lack of studies examining individuals with an impulse-control disorder, the current review will focus on other psychiatric conditions characterized, at least partially, by impulsive behaviors, primarily ADHD, mood disorders, and addiction.

3.1. Glutamatergic Genes Implicated in Disorders Characterized by Impulsive Behaviors

The GRIN2B gene, which encodes for the GluN2B subunit of the NMDA receptor, is linked to both ADHD and bipolar disorder (Dorval et al., 2007; Kim et al., 2018; Zhao et al., 2011), and long-term use of alcohol or cocaine leads to upregulated GRIN2B expression in the brain (Enoch et al., 2014). Specifically, the T-C-C haplotype of the GRIN2B gene is transmitted more frequently in individuals with bipolar disorder (Martucci et al., 2006). However, two GRIN2B single nucleotide polymorphisms (SNPs) (rs2300256 and rs2284411) are associated with inattentive symptoms as reported by one’s parent and/or teacher (Dorval et al., 2007; but see Park et al., 2013), while one additional SNP of the GRIN2B gene (rs2284407) is only marginally associated with hyperactive/impulsive symptoms as reported by parents (Dorval et al., 2007). These results suggest that the GluN2B subunit may be more involved in attentional deficits as opposed to impulsive behavior in individuals with ADHD. In addition to the GRIN2B gene, SNPs of ionotropic glutamate receptors are associated with psychiatric disorders characterized, at least in part, by impulsive behaviors. The GRIN2A gene is associated with ADHD (Turic et al., 2004; but see Adams et al., 2004) and heroin use disorder, at least in in African Americans (Levran et al., 2009). SNPs for the CACNG8 gene are linked to ADHD (Bai et al., 2022) and antisocial personality disorder (Peng et al., 2021), and genes encoding kainate receptors are associated with ADHD (Chatterjee et al., 2022) and aggressive behaviors in individuals with bipolar disorder (Ma et al., 2019). Overall, modifications to excitatory receptors seem to serve as a risk factor for the emergence of maladaptive impulsive behaviors.

Genetic association studies also implicate mGluRs in conditions like ADHD and mood disorders. Specifically, SNPs for GRM7 (Noroozi et al., 2019; Zhang et al., 2021a; but see Akutagava-Martins et al., 2014), as well as deletions of copy variants of genes encoding for mGluR₁, mGluR₅, mGluR₇, and mGluR₈, are associated with ADHD (Glessner et al., 2023; Liu et al., 2021), whereas SNPs for the GRM4 gene are protective against ADHD (Zhang et al., 2021a). Like ADHD, two SNPs for the GRM7 gene are observed more frequently in individuals with MDD and bipolar II disorder (Noroozi et al., 2019), and copy number variants of two additional GRM7 SNPs are linked to bipolar disorder (Kandaswamy et al., 2014).

Although the results summarized above implicate glutamatergic genes to psychiatric conditions related to impulsive behaviors, one limitation of genetic association studies is that they do not elucidate how glutamatergic dysfunction mediates such behavior. Relatedly, differential gene expression does not guarantee that protein expression and/or function will be altered. Due to this limitation, studies measuring glutamate/Glx levels, as well as mRNA/protein expression, can provide valuable information concerning aberrant glutamatergic systems in individuals with a psychiatric disorder linked to impulsivity. Below, I will provide evidence of glutamatergic alterations observed in individuals with a psychiatric disorder characterized by impulsive behaviors. I will begin by discussing glutamatergic alterations in the peripheral nervous system.

3.2. Glutamatergic Alterations in Peripheral Nervous System

Peripheral nervous system glutamate/Glx levels and mRNA expression can be quantified from one’s blood plasma. Blood plasma glutamate levels are elevated in children with ADHD, individuals with MDD, and individuals with substance use disorders, with corresponding decreased glutamine levels observed in these conditions (Inoshita et al., 2018; Nishi et al., 2015; Ogawa et al., 2018; Sari et al., 2020; Skalny et al., 2021; Umehara et al., 2017; Wang et al., 2021; Wei et al., 2020). Additionally, peripheral glutamate receptors are more sensitive to glutamate stimulation in individuals with MDD (Berk et al., 2001). One potential explanation for increased peripheral glutamate observed in these individuals is deficient activity of the enzyme glutamine synthetase, leading to overexcitation of glutamatergic receptors. However, this is speculative as glutamine synthetase activity in peripheral nervous system has not been directly measured in individuals with impulse-control related disorders. Another possibility is deficient activity of glutamate decarboxylase (GAD), the enzyme responsible for converting glutamate into the major inhibitory neurotransmitter gamma amino-butyric acid (GABA). There is some evidence to support this hypothesis; individuals with the rs3749034 G allele or the rs11542313 C allele of the GAD1 gene score higher in measures of hyperactivity and impulsivity (Bruxel et al., 2016), and decreased peripheral GAD1 mRNA expression is observed in individuals with MDD (Lin et al., 2019). In addition to deficient GAD enzyme activity, decreased functionality of EAATs may contribute to increased peripheral glutamate levels as the SLC1A3 gene is implicated in ADHD (Laurin et al., 2008; van Amen-Hellebrekers et al., 2016).

In contrast to ADHD, MDD, and substance use disorders, glutamine serum levels are elevated in individuals with bipolar disorder (Pålsson et al., 2015), and glutamate levels are lower in individuals with bipolar disorder experiencing their first psychotic episode (Palomino et al., 2007). Lower glutamate levels are also observed in men with internet gaming disorder (Paik et al., 2018). Individuals experiencing acute mania have elevated antibody levels for the GluN2 subunit of the NMDA receptor (Dickerson et al., 2012), perhaps reflecting a compensatory upregulation in response to decreased glutamate serum levels. The decreased glutamate levels may stem from impaired glutaminase activity in peripheral nervous system and may reflect an important mechanism that distinguishes MDD from bipolar disorder and substance use disorders from behavioral addictions. Research is needed to directly test this hypothesis.

The altered glutamatergic activity observed in individuals with ADHD and mood disorders does not, on its own, provide an explanation for increased impulsive behaviors associated with these conditions. Glutamate is an important mediator of the somatic nervous system, as sensory neurons that transmit information regarding nociception, vision, audition, olfaction, and taste rely on glutamatergic signaling (Kalloniatis et al., 1996; Miller et al., 2011; Murphy et al., 2004; Vandenbeuch and Kinnamon, 2016; Weisz et al., 2021). However, glutamatergic control of motor neurons is primarily localized to the central nervous system, with acetylcholine innervating skeletal muscles in the peripheral nervous system (Stifani, 2014). As such, a discussion of glutamatergic disturbances in the central nervous system will be presented next.

3.3. Glutamatergic Alterations in Cerebrospinal Fluid

One way to measure glutamatergic activity in the central nervous system is to quantify glutamate/Glx levels in cerebrospinal fluid (CSF). Because this approach does not elucidate where in the brain aberrant glutamatergic neurotransmission occurs, relatively few studies focus exclusively on measuring Glx concentrations in CSF. Indeed, examination of Glx levels in CSF has largely been confined to individuals with mood disorders. To this end, no differences in Glx levels are observed in patients with MDD or bipolar disorder compared to controls (Garakani et al., 2013; Hashimoto et al., 2016; Pålsson et al., 2015). Yet, the glutamine-to-glutamate ratio is higher in individuals with MDD, suggesting an abnormal glutamine-glutamate cycle (Hashimoto et al., 2016). This finding is interesting as it suggests the opposite relationship between glutamine and glutamate levels in CSF (increased glutamine and decreased glutamate) compared to the peripheral nervous system (decreased glutamine and increased glutamate). Additional research shows increased glutamine levels in CSF of individuals with MDD (Levine et al., 2000), and individuals with refractory depressive disorders (i.e., those with MDD or bipolar disorder that do not respond to various pharmacotherapies) have lower glutamate levels (Frye et al., 2007a).

Although CSF glutamate levels are lower in individuals with depression, glutamate levels are positively correlated with reactive aggression and suicidal ideation in this population (Coccaro et al., 2013; Garakani et al., 2013), indicating that increased glutamate neurotransmission can increase impulsive-like behaviors in those diagnosed with a mood disorder. Additionally, quinolinic acid, an NMDA receptor agonist, is elevated in individuals that have previously attempted suicide, with quinolinic acid levels positively correlating with suicidal ideation (Erhardt et al., 2013). The positive association observed between CSF glutamate levels and aggression is consistent with increased blood glutamate in individuals with antisocial personality disorder (Gulsun et al., 2016). Preclinical research supports this hypothesis as GAD-deficient mice and GLUT-3 knockout mice are more hyperactive compared to controls (Shin et al., 2021; Smith, 2018).

Like mood disorders, decreased CSF glutamate levels are observed in individuals classified as heavy smokers, with these individuals scoring higher in motor impulsivity as assessed with the BIS-11 (Li et al., 2018). One concerning finding is that alcohol withdrawal leads to an increase in CSF glutamate concentrations (Tsai et al., 1998), as well as in peripheral glutamate concentrations (Aliyev et al., 1994). Preclinical models show increased impulsive behaviors during alcohol withdrawal (Irimia et al., 2014; Irimia et al., 2015; Irimia et al., 2017), providing additional support that aberrant glutamatergic transmission exacerbates impulsivity.

The discrepancy in glutamine/glutamate ratios in peripheral and central nervous systems for individuals with a condition like ADHD or MDD may result from a weakened blood-brain barrier. A dysfunctional blood-brain barrier is observed in individuals with MDD (Wu et al., 2022), and neuroinflammation, which can weaken the blood-brain barrier (Galea, 2021), is proposed to act as a risk factor for ADHD (Dunn et al., 2019). Glutamate released in the central nervous system may “leak” through a weakened blood-brain barrier into the peripheral nervous system. The net result would be elevated glutamate levels in peripheral nervous system and lower glutamate levels in CSF. Tangentially related to this point, increased glutamate consumption decreases the odds of being diagnosed with ADHD during childhood whereas increased consumption of aspartate and glycine increases the odds of being diagnosed with ADHD (Holton et al., 2019). By consuming foods high in glutamate, at-risk children for ADHD may be able to replenish glutamate concentrations in the central nervous system, thus preventing the expression of ADHD-associated impulsive behaviors. Interestingly, the significant association between amino acid consumption and ADHD is specific to children/young adolescents, as this association disappears in college-aged individuals (Holton et al., 2019). Children at risk of ADHD may have a weakened blood-brain barrier that becomes stronger over time. This may provide an explanation as to why ADHD is primarily first diagnosed during childhood, with ADHD diagnoses decreasing across the lifespan (Vos and Hartman, 2022).

3.4. Glutamatergic Control of Impulsive Behaviors: Evidence from Behavioral Pharmacology Experiments

Instead of correlating CSF glutamate levels to impulsive behaviors, behavioral pharmacology experiments are used to determine how administration of a glutamatergic ligand affects impulsivity. Given that elevated CSF glutamate levels are associated with affective impulsivity and drug withdrawal-induced impulsivity (see section above), decreasing excitatory neurotransmission should be effective in attenuating impulsive behaviors. To this end, anticonvulsant medications directly inhibit excitatory neurotransmission by either blocking sodium channels or by enhancing GABA function (Stafstrom, 2010; Zanatta et al., 2019), leading to decreased glutamate concentrations (Soeiro-de-Souza et al., 2018) with corresponding increased glutamine levels (Soeiro-de-Souza et al., 2015). Specific to impulsive behavior, anticonvulsants, particularly diavproex (valproate), reduce impulsive aggression in individuals with or at risk for bipolar disorder (Barzman et al., 2005; Barzman et al., 2006; Chang et al., 2006; MacMillan et al., 2006; Saxena et al., 2006), as well as in children with ADHD (Blader et al., 2009) and in individuals with borderline personality disorder (Hollander et al., 2005; Kavoussi and Coccaro, 1998; Tritt et al., 2005). Similarly, the anticonvulsant lamotrigine decreases aggressive behavior in rats (Kumar et al., 2016).

The results described above support the hypothesis that increased glutamatergic transmission leads to increased impulsivity. To better understand the molecular underpinnings of glutamatergic-mediated impulsivity, preclinical behavioral pharmacology experiments and clinical trials can use ligands that selectively bind to a molecular target. These include ionotropic glutamate receptor antagonists, Group I mGluR antagonists, Group II mGluR agonists, and allosteric modulators of glutamate receptors. Allosteric modulators bind to a different site on the receptor compared to agonists and change the receptor’s response to agonists. Positive allosteric modulators increase the affinity and/or efficacy of an agonist while negative allosteric modulators decrease the affinity and/or efficacy of an agonist (see Conn et al., 2009 for a general review of allosteric modulation; see Gasparini and Spooren, 2007 for a review of mGluR allosteric modulators; see Brogi et al., 2019 for a review of ionotropic glutamate receptor allosteric modulators).

Concerning ionotropic glutamate receptors, few studies focus on the effects of AMPA receptor ligands. While the AMPA receptor antagonist CNQX fails to alter impulsive choice (Yates et al., 2015), an AMPA receptor positive allosteric modulator reduces motor impulsivity in the 1CSRTT (Davis-Reyes et al., 2021), which is consistent with the reduced impulsivity observed in individuals with ADHD following treatment with an AMPA receptor positive allosteric modulator (Adler et al., 2012). In contrast to the AMPA receptor, studies examining the effects of NMDA receptor ligands on impulsive behaviors are more common. Antagonists of NMDA receptors can either (a) block the channel opening (i.e., uncompetitive receptor antagonist), (b) bind to the same site as the agonist (i.e., competitive receptor antagonist), or (c) target NMDA receptors that contain a specific subunit (primarily GluN2B-selective antagonists). Some commonly known drugs act as NMDA receptor channel blockers, including ketamine, memantine, and phencyclidine. One seemingly paradoxical finding is that NMDA receptor antagonists increase glutamate concentrations throughout the brain (Abdallah et al., 2018; Amitai et al., 2012; Ito et al., 2006; Léna et al., 2007; Lisek et al., 2017; Moghaddam et al., 1997; Razoux et al., 2007; Zuo et al., 2006). The increased glutamatergic activity following NMDA receptor antagonism is most likely due the inhibition of inhibitory interneurons (e.g., Hudson et al., 2020; see Lee and Zhou, 2019). As such, NMDA receptor antagonists, including GluN2B-selective antagonists, frequently increase locomotor hyperactivity (Fredriksson and Archer, 2003; Higgins et al., 2016; Popik et al., 2015) and increase motor impulsivity in outbred rodents across a variety of tasks (Burton and Fletcher, 2012; Carli et al., 2011; Higgins et al., 2016; Jentsch and Anzivino, 2004; Terry Jr. et al., 2012; see Table 2 for citations for specific tasks).

Table 2.

Summary of Results of Preclinical Behavioral Pharmacology Experiments Examining Contributions of Glutamate to Motor Impulsivity and Impulsive Choice

Molecular Target	Agonist/PAM	Antagonist/NAM	Region (if not systemically administered)	Motor Impulsivity (Specific task)	Impulsive Choice	References
mGluR 1
		JNJ16567083 (EMQMCM)		↓ (DRL)	↓	Sukhotina et al. (2008)
		JNJ16259685			↑	Yates et al. (2017b)
mGluR 5
		RO4917523		↓ (5CSRTT)	--- (5CSRTT)	Isherwood et al. (2015)
		MTEP		NS (5CSRTT)	--- (5CSRTT)	Isherwood et al. (2015)
		MPEP		↓ (5CSRTT)		Semenova & Markou (2007)
					---	Yates et al. (2017b)
	ADX47273			↓ (5CSRTT)		Liu et al. (2008)
				--- (5CSRTT)		Isherwood et al. (2015)
mGluR 2/3
		LY341495		--- (5CSRTT)		Semenova & Markou (2007)
	LY379268			↓ (5CSRTT)		van der Veen et al. (2021)
mGluR 2
	TASP0433864			↓ (5CSRTT)		van der Veen et al. (2021)
mGluR 4
	Cpd11			↑ (Go/no-goa)		Piszczek et al. (2022)
				↑ (5CSRTT)		Isherwood et al. (2017)
AMPA Receptor
	HJC0122			↓ (1CSRTT)		Davis-Reyes et al. (2021)
		CNQX	NAc & systemic		---	Yates and Bardo (2017); Yates et al. (2015)
NMDA Receptor
NMDA receptor channel blockers		Phencyclidine (PCP)		↑ (lateralized reaction time task)		Jentsch & Anzivino (2004)
				↑ (5CSRTT)		Barnes et al. (2018)
				---- (5CSRTT)b		Benn & Robinson (2014)
		MK-801 (dizocilipne)		↑ (5CSRTT, go/no-go, & rat gambling task)		Benn & Robinson (2014); Higgins et al. (2016); Higgins et al. (2018) Isherwood et al. (2015); Onofrychuk et al. (2021); Popik et al. (2015); Terry Jr. et al. (2012)
				NS (5CSRTT)		Higgins et al. (2005)
					↓
					---	Higgins et al. (2016); Yates et al. (2015); Yates et al. (2019)
					NS	Yates et al. (2017a)_
						Higgins et al. (2018)
		Ketamine		↓ (1CSRTT)c		Davis-Reyes et al. (2021)
				NS (5CSRTT)		Benn & Robinson (2014)
					↑	Cottone et al. (2013)
					NS	Yates et al. (2015); Yates et al. (2017a)
		Memantine		NS (5CSRTT)		Benn & Robinson (2014)
						Cottone et al. (2013)
					↑	Yates et al. (2017a)
					NS
NMDA receptor competitive antagonists		AP-5	NAc & systemic		---	Cottone et al. (2013); Yates and Bardo (2015)
		CGS 19755			↑	Cottone et al. (2013); Yates et al. (2017a)
		CPP	mPFC	↑ (5CSRTT)		Carli et al. (2011)
GluN2B-selective antagonists		Ifenprodil	NAc & systemic	--- (5CSRTT)		Higgins et al. (2005)
						Yates & Bardo (2017)
					↓ (when injected into NAc)	Yates et al. (2017a)
					NS (when injected systemically)
		Ro 25-6981		↑ (5CSRTT)		Higgins et al. (2005)
		Ro 63-1908		↑ (2CSRTT, 5CSRTT, go/no-go, & rat gambling task)		Burton & Fletcher, 2012; Higgins et al. (2005); Higgins et al. (2016); Higgins et al. (2018); van der Veen et al. (2021)
					↑	Higgins et al. (2016); Yates et al. (2018)
		CP-101,606 (traxoprodil)		↑ (5CSRTT & go/no-go)		Higgins et al. (2005); Higgins et al. (2016)
					---	Higgins et al. (2016); Yates et al. (2018)
Restoration of glutamate homeostasis via N-acetylcysteine				↓ (5CSRTT)		Fredricksson et al. (2023)

Abbreviations: AP-5 = (2R)-amino-5-phosphonovaleric acid. CGS 19755 = cis-4-phosphonomethyl-2-piperidine carboxylic acid. CNQX = 6-cyano-7-nitroquinoxaline-2,3-dione. CPP = 3-((+/−)-2-carboxypiperazin-4-yl)propyl-1-phosphonic acid. CSRTT = choice serial reaction time task. DRL = differential reinforcement of low rates of responding. MPEP = 2-Methyl-6-(phenylethynyl)pyridine. MTEP = 3-((2-Methyl-4-thiazolyl)ethynyl)pyridine. NAM = negative allosteric modulator. NS = non-selective drug effects (e.g., increased omissions, response latencies, decreased preference for large magnitude reinforcer when delay is set to 0 s that make interpreting drug-induced changes in impulsive behavior difficult). PAM = positive allosteric modulator.

Notes. The design of the table is based on Table 1 from Pattij and Vanderschuren (2008).

^aIn low impulsive rats only.

^bThere appears to be some increases in motor impulsivity at 1.0 mg/kg, but this effect disappears at a higher dose (3.0 mg/kg).

^cThere was a trend for increased omissions.

Compared to motor impulsivity, the effects of NMDA receptor antagonists on impulsive choice are more nuanced, appearing to be dependent on the specific ligand used and the design of the delay-discounting task. The NMDA receptor competitive antagonist AP-5 fails to alter delay discounting (Cottone et al., 2013; Yates and Bardo, 2017), but the NMDA receptor competitive antagonist CGS 19755 increases impulsive choice (Cottone et al., 2013; Yates et al., 2017a). Similar discrepancies are observed for GluN2B-selective antagonists. Although Ro 63–1908 increases impulsive choice, CP-101,606 does not affect impulsive choice (Yates et al., 2018). Furthermore, ifenprodil leads to decreased preference for a large magnitude reinforcer when its delivery is immediate without selectively altering impulsive choice (Yates et al., 2017a). One of the major challenges of preclinical delay-discounting procedures is dissociating impulsive choice from other constructs such as behavioral flexibility and suboptimal choice (see Yates, 2018 for a full discussion). For example, when the delay to receiving the large, delayed reinforcer systematically increases across the session, the NMDA receptor uncompetitive antagonist MK-801 decreases impulsive choice (Higgins et al., 2018; Yates et al., 2015; Yates et al., 2019; but see Yates et al., 2017a for null effects), which contradicts the effects of NMDA receptor antagonists on motor impulsivity; yet, MK-801 increases impulsive choice when the delay to this alternative decreases across a session (Higgins et al., 2018). This finding indicates that MK-801 increases perseverant responding in this task without necessarily altering impulsive choice. Ketamine and memantine are reported to increase impulsive choice in an adjusting delay procedure (Cottone et al., 2013), but these drugs decrease preference for a large magnitude reinforcer even when the delay to its delivery is set to 0 s (Yates et al., 2015; Yates et al., 2017a), indicating increased suboptimal choice. Thus, the changes in behavior observed following glutamatergic manipulations may not indicate selective alterations in impulsive choice. As such, more work is needed to isolate the influence of glutamate on impulsive choice from behavioral flexibility in preclinical models of delay discounting.

Whereas NMDA receptor antagonists increase motor impulsivity in animals, the NMDA receptor channel blocker memantine improves both hyperactive and inattentive symptoms in adults with ADHD (Surman et al., 2013) and improves total ADHD symptoms as assessed with the ADHD Rating Scale-IV in children, although the effects of memantine on each domain of ADHD are modest (Findling et al., 2007). Ketamine provides rapid symptom improvements for those with MDD (Berman et al., 2000; Fava et al., 2020; Ghasemi et al., 2014; Mandal et al., 2019; Murrough et al., 2013); directly related to the current review, ketamine reduces suicidal ideation (Ballard et al., 2014; Murrough et al., 2015). The discrepancies observed between preclinical and clinical studies most likely stems from the use of outbred rodent strains that do not model aspects of psychiatric conditions. Similar discrepancies are observed for ADHD treatments like d-amphetamine and methylphenidate. These drugs are effective at reducing impulsive behaviors in individuals with ADHD, but they can increase motor impulsivity in healthy individuals (e.g., Voon et al., 2015).

Related to this point, the ability of NMDA receptor antagonists to improve impulsivity in clinical populations may be related to glutamate homeostasis. Glutamate homeostasis is already implicated in substance use disorders (see Kalivas, 2009), and there is increased focus on using N-acetylcysteine, which restores glutamate homeostasis by activating the x_C⁻ system, for the treatment of various psychiatric disorders (Smaga et al., 2021), although there are few studies testing the effects of N-acetylcysteine on impulsive behaviors. Although not directly tested in individuals with ADHD, N-acetylcysteine improves ADHD Self-Report Scale scores in individuals with lupus (Garcia et al., 2013) and ameliorates motor impulsivity (Fredriksson et al., 2023). The ability of N-acetylcysteine to decrease impulsivity most likely results from increased extrasynaptic glutamate that binds to Group II/III mGluRs on the presynaptic neuron, thus regulating the further release of glutamate into the synapse (see Smaga et al., 2021 for a review). Further supporting this argument is that mGluR_2/3 agonists and mGluR₂ positive allosteric modulators reduce premature responses in the 5CSRT (van der Veen et al., 2021) and attenuate serotonergic- and glutamatergic-induced increases in motor impulsivity (Pozzi et al., 2011; van der Veen et al., 2021; Wischhof and Koch, 2012; Wischhof et al., 2011) while mGluR₂ knockout mice exhibit enhanced locomotor activity (Morishima et al., 2005). The synthetic molecule fasoracetam, which acts as a non-selective mGluR activator, improves ADHD symptom severity in adolescents (Elia et al., 2018). Conversely, mGluR₅ antagonists typically fail to alter motor impulsivity and impulsive choice (Isherwood et al., 2015; Semenova and Markou, 2007; Yates et al., 2017b; but see Liu et al., 2008) while mGluR₁ antagonists increase motor impulsivity (Sukhotina et al., 2008) and impulsive choice (Yates et al., 2017b; but see Sukhotina et al., 2008).

In contrast to mGluR₂ positive allosteric modulators, mGluR₄ positive allosteric modulators increase motor impulsivity in mice that exhibit low baseline levels of motor impulsivity in a go/no-go task (Piszczek et al., 2022) and in rats in a 5CSRTT (Isherwood et al., 2017). The discrepancies observed between mGluR₂ positive allosteric modulators and mGluR₄ positive allosteric modulators may be related to regional differences between these receptor subtypes. mGluR₂ mRNA is primarily expressed in the dentate gyrus of the hippocampus and in the cerebellum while mGluR₄ expression is highest in cerebellum and in ventrolateral/ventromedial nuclei of the thalamus; mGluR₄ mRNA is also higher in parts of the striatum relative to mGluR₂ (Testa et al., 1994). Although speculative, these regional differences emphasize the importance of elucidating the neural circuitry involved in glutamatergic control of impulsive behaviors. This is the topic of the next section.

3.5. Glutamatergic Pathways Regulating Impulsive Behaviors

Figure 2 shows glutamatergic projections throughout the brain. There are multiple glutamatergic pathways that originate from the prefrontal cortex (PFC) (see Gasiorowska et al., 2021; Schwartz et al., 2012). The PFC is the most anterior region of the brain and is typically subdivided into lateral PFC (lPFC) and ventromedial PFC (vmPFC), with these regions being further subdivided into dorsolateral PFC (dlPFC), ventrolateral PFC (vlPFC), medial PFC (mPFC), and ventral PFC (vPFC). The orbitofrontal cortex (OFC) is also part of the PFC and is subdivided into medial OFC (mOFC) and lateral OFC (lOFC). Collectively, the structures of the PFC are involved in processes such as language, memory, and executive function (Siddiqui et al., 2008), including decision making (Euston et al., 2012; Hiser and Koenigs, 2018). Not surprisingly, the PFC is a critical mediator of impulsive behaviors, with subregions of the PFC making differential contributions to impulsivity constructs (Kim and Lee, 2010).

Figure 2.

A schematic depicting glutamatergic projections in the human brain. Abbreviations: ACC = anterior cingulate cortex, Amy = amygdala, DRN = dorsal raphe nuclei, Hip = hippocampus, NAc = nucleus accumbens, PFC = prefrontal cortex, SMA = supplemental motor area, STN = subthalmic nucleus, STR = striatum, TH = thalamus, VTA = ventral tegmental area. Created with BioRender.com.

PFC glutamatergic neurons innervate basal ganglia, specifically dorsal striatum, cerebellum, and motor cortical areas located in the posterior frontal cortex (supplemental motor area, premotor cortex, and primary motor cortex). The basal ganglia are subcortical structures that help facilitate purposeful movement while inhibiting interfering movements (Utter and Basso, 2008) and mediate habit formation (Yin and Knowlton, 2006). Historically believed to regulate fine motor control and balance (Ito, 1984), the cerebellum is now theorized to be involved in multiple higher-order cognitive processes such as executive control and emotional and social processing (Beuriat et al., 2022). Collectively, the motor cortices plan and initiate movements (Nachev et al., 2008; Roland, 1984).

Structures in the limbic system, which are involved in memory, emotion, and motivation, also receive glutamatergic inputs from PFC. The most prominent structures of the limbic system are hippocampus, amygdala, and anterior cingulate cortex (ACC) (Morgane et al., 2005). A critical function of the hippocampus is consolidating short-term memory into long-term memory (Squire et al., 2015). The amygdala is involved in emotional processing (Sergerie et al., 2008), especially fear (Davis, 1992). The ACC is implicated in multiple processes, including conflict monitoring, attention allocation, and decision making (Rolls, 2019). The nucleus accumbens (NAc), a region located in the ventral striatum, is closely related to the limbic system and plays a critical role in reward learning (Day and Carelli, 2007) and the development of addictions (Scofield et al., 2016).

Glutamatergic neurons originating from PFC travel to multiple nuclei located in the brainstem. For the purpose of the current review, I will briefly mention two relevant structures: the substantia nigra and the ventral tegmental area (VTA). The substantia nigra sends dopaminergic neurons to basal ganglia structures, thus serving as a critical mediator of motor control (Luo and Huang, 2016). Stimulation of VTA leads to dopamine release in NAc (Chen, 2022). The final two major glutamate pathways involve the reciprocal connection between PFC and thalamus. The thalamus has numerous functions, such as relaying sensory information to other regions of the brain, regulating sleep, and modulating consciousness (Biesbroek et al., 2024).

Because prefrontal glutamatergic neurons are part of motor control circuits and innervate structures of the limbic system, dysregulation of intra-PFC should be a major contributor of distinct forms of impulsivity. The section below will focus exclusively on aberrations in prefrontal glutamatergic activity that are associated with impulsive behaviors. As impulsivity measures often do not correlate with one another in both humans and animals (e.g., Broos et al., 2012; Marusich et al., 2011; see Strickland and Johnson, 2021 for a review), discussion of how differential dysregulation of glutamatergic pathways accounts for the different types of impulsivity is crucial.

3.5.1. Evidence that Glutamatergic Dysfunction in PFC Mediates Impulsive Behaviors.

Altered prefrontal glutamate levels are often observed in individuals with ADHD, mood disorders, and substance use disorders. Increased glutamate levels in right dlPFC (Bakhshi et al., 2022; MacMaster et al., 2003; but see Carrey et al., 2007), but not in left dlPFC (Bollmann et al., 2015; Puts et al., 2020) are observed in ADHD children, although decreased glutamate levels in right PFC have been reported (Hai et al., 2020a). These inconsistencies may be due to ADHD subtype. For example, children diagnosed with the hyperactive subtype, but not the inattentive subtype, of ADHD have elevated Glx levels in frontal cortex (Courvoisie et al., 2004) while adults primarily diagnosed with the inattentive or the combined subtypes of ADHD have decreased glutamate levels in left midfrontal region (Dramsdahl et al., 2011). When healthy participants are used, there is a positive association between glutamatergic activity in dlPFC and increased motor impulsivity in an auditory go/no-go task (Koizumi et al., 2018). Although individuals with a genetic marker for MDD or bipolar disorder have lower dlPFC glutamate levels (Thomson et al., 2016), increased glutamate/glutamine levels are observed in PFC of individuals diagnosed with bipolar disorder (Castillo et al., 2000; Lan et al., 2009; Michael et al., 2009; but see Smaragdi et al., 2019). Similarly, elevated glutamate levels are observed in the dlPFC of individuals diagnosed with a Cluster B personality disorder, which includes borderline personality disorder and antisocial personality disorder (Smesny et al., 2018). Drug dependence is often associated with elevated glutamate levels in mPFC (Li et al., 2020). Importantly, glutamate levels in mPFC positively correlate with self-reported impulsivity measures in dependent individuals (Li et al., 2020) while glutamate levels in left dlPFC are correlated with cravings for alcohol (Frye et al., 2016). The elevated glutamate observed in mPFC may partially account for why individuals have difficulty inhibiting themselves from using a substance even in the face of negative consequences (e.g., potential for overdose, health complications, legal issues). Indeed, drug craving during abstinence is associated with impulsivity (Doran et al., 2009; Evren et al., 2012; Roozen et al., 2011).

Animal research supports the hypothesis that motor impulsivity is influenced by hyperactive prefrontal glutamatergic activity. Increased excitatory neurotransmitter levels are observed in SHRs in PFC, with evoked glutamate release and glutamate uptake being observed as well (Miller et al., 2014; Rizzo et al., 2017). While SHRs have decreased tonic glutamate signaling in the frontal cortex compared to WKYs, they show increased phasic glutamate signaling (Miller et al., 2019; but see Miller et al., 2014 for a non-statistically significant increase in tonic glutamate signaling in PFC of SHRs), and they show increased glutamate-stimulated release of norepinephrine (Russell, 2001). Intracellular Ca⁺⁺ levels are elevated in prefrontal cortical neurons of SHRs (Lehohla et al., 2004), which may account for the increased glutamate release observed in these rats. Furthermore, SHRs have increased CAMKII activity in mPFC, which leads to upregulated GluR1 subunit levels (Yabuki et al., 2014). Infusion of NMDA receptor antagonists into infralimbic, but not prelimbic, region of mPFC increases motor impulsivity (Agnoli and Carli, 2012; Benn and Robinson, 2014; Carli et al., 2006; Murphy et al., 2005; Murphy et al., 2012; Pozzi et al., 2011), and preventing the conversion of glutamate to GABA in mPFC increases hyperactivity and increases premature responses in the 5CSRTT (Asinof and Paine, 2013).

Like motor impulsivity, there is evidence that a hyperactive prefrontal glutamatergic system influences affective impulsivity. Increased Glx levels in dlPFC are positively correlated with aggression in individuals with antisocial personality disorder (Smaragdi et al., 2019). Increased Glx levels may be attributed to SNPs in the GAD1 gene that lead to reduced GAD expression, thus leading to accumulating glutamate levels in PFC (Arrúe et al., 2019). Indeed, individuals with bipolar disorder have lower GAD-67 mRNA expression in PFC (Guidotti et al., 2000; Thompson et al., 2009), and they have decreased full-length transcripts, but increased truncated transcripts, of the GAD2 gene in dlPFC (Davis et al., 2016). In addition to altered GAD expression, increased VGLUT-1 and VGLUT-2 expression in dlPFC is observed in women who died by suicide (Powers et al., 2020). Increased VGLUT expression suggests that excess glutamate is being stored in vesicles to be released into the synapse. Men that died by violent suicide have lower EAAT-1 expression than those that died by non-violent suicide, with a similar trend observed for EAAT-2 expression (Powers et al., 2020). These results are consistent with the finding that CSF glutamate levels are positively correlated with suicidal ideations (Garakani et al., 2013). Individuals that have died by suicide, regardless of gender, have increased AMPA receptor (GluR2–4), kainate receptor (subunits 1 and 3), GluN1, GluN2A, GluN2B, mGluR₁, mGluR₂, and PSD mRNA expression in dlPFC (Gray et al., 2015; Zhao et al., 2018), providing further support for a hyperactive glutamatergic system in individuals prone to engaging in forms of affective impulsivity.

In contrast to motor impulsivity and affective impulsivity, a hypoactive prefrontal glutamatergic system appears to mediate impulsive choice, although the evidence supporting this hypothesis is more indirect compared to other forms of impulsivity. Decreased dlPFC activation is observed in individuals with ADHD during delay-discounting performance (Ortiz et al., 2015), and delay aversion-induced increases in dlPFC activation are observed in male adolescents with ADHD (Van Dessel et al., 2018). The greater activation of dlPFC observed in adolescents with ADHD, relative to controls, may be due to lower baseline levels of excitatory neurotransmission. Children with ADHD, primarily combined subtype, show improvements in impulsive choice following concurrent anodal transcranial direct stimulation of right vmPFC and cathodal stimulation of left dlPFC (Nejati et al., 2021). Specific to bipolar disorder, manic symptoms and depressive symptoms are correlated with BIS-11 attentional impulsivity (Swann et al., 2008). However, mania is positively associated with motor impulsivity while hopelessness is correlated with non-planning impulsivity (Swann et al., 2007; Swann et al., 2008). This result is interesting as non-planning impulsivity is correlated with delay discounting (Baumann and Odum, 2012), and individuals with MDD often show increased impulsive choice in delay-discounting tasks (e.g., Takahashi et al., 2008). Perhaps the lower intra-PFC glutamate levels commonly observed in individuals with MDD (e.g., Ritter et al., 2022; Shirayama et al., 2017) account for increased non-planning impulsivity/impulsive choice.

If hyperactive glutamatergic signaling in PFC contributes to motor/affective impulsivity, the ability of drugs that increase glutamate levels in regions like PFC to improve ADHD symptoms (e.g., memantine; see Section 3.4.) seems paradoxical. One potential explanation to reconcile this discrepancy is that increasing glutamate transmission in PFC restores glutamate homeostasis by allowing glutamate to more easily bind to Group II/Group mGluRs, which regulate the further synthesis and the release of glutamate, thus normalizing glutamate levels in high impulsive individuals. Preclinical research shows that expression of mGluR₂ is lower in older adult rats compared to younger adult rats (Hernandez et al., 2018) and decreases in the cortex of mice across development (McOmish et al., 2016). As synaptic pruning of excitatory synapses occurs between adolescence and adulthood (Selemon, 2013), administration of a drug like memantine may allow for glutamate to directly stimulate postsynaptic glutamate receptors, resulting in amelioration of impulsive behaviors in adults with ADHD. This hypothesis is supported by the finding that rats that display increased motor impulsivity have lower mGluR₂ mRNA and protein expression and lower mGluR₂ receptor binding in mPFC (Elfving et al., 2019; Fomsgaard et al., 2018; Klein et al., 2014). Somewhat relatedly, although systemic administration of lamotrigine decreases reactive aggression (Kumar et al., 2016), lamotrigine fails to ameliorate premature responses following NMDA receptor antagonist infusion into infralimbic cortex (Murphy et al., 2012); in fact, long-term administration of lamotrigine into mPFC increases motor impulsivity (Paine et al., 2015). Long-term lamotrigine administration may lead to a loss of glutamate homeostasis as insufficient glutamate release can prevent the stimulation of extrasynaptic inhibitory mGluRs.

Loss of glutamate homeostasis may also account for altered receptor mRNA/protein expression observed in PFC of high impulsive animals. Despite upregulated GluR1 expression, SHRs have decreased expression of AMPA receptors and reduced AMPA receptor-mediated neurotransmission, with unaltered NMDA receptor-mediated neurotransmission, in PFC (Cheng et al., 2017). Conversely, Roman high avoidance (RHA) rats, a strain of rat selectively bred from Wistar rats (Bignami, 1965) that display enhanced motor impulsivity (Arrondeau et al., 2023; Bellés et al., 2021; Klein et al., 2014), have elevated levels of scaffolding proteins and increased mRNA expression for the GluN2B subunit in PFC compared to Roman low avoidance (RLA) rats (Elfving et al., 2019). Likewise, non-ADHD-like rats demonstrating high motor impulsivity have lower GluN1 and GluN2A, but higher GluN2B, expression in mPFC relative to low impulsive rats (Davis-Reyes et al., 2019). These results are interesting as reduced AMPA receptor-mediated excitatory neurotransmission and increased GluN2B subunit expression indicate weakened synaptic strength (e.g., Kim et al., 2005). At first glance, reduced synaptic strength observed in PFC would suggest a hypoactive glutamatergic system in this region of high impulsive individuals. Indeed, hypoactive PFC is argued to underlie ADHD (see Cheng et al., 2017 for an example). Yet, a hyperactive glutamatergic system in PFC is not necessarily at odds with the hypothesis of hypoactive PFC functioning. Hyperactive glutamatergic signaling can lead to over excitation of inhibitory interneurons that are abundant in the PFC (Ferguson and Gao, 2018), leading to a net result of hypoactive activity. Additionally, excess glutamate causes AMPA receptor internalization (Zhou et al., 2001), which leads to a weakened synapse (Shepherd and Huganir, 2007).

3.5.2. Glutamate and Motor Impulsivity.

Figure 3a shows the glutamatergic pathways implicated in motor impulsivity. The selection and the inhibition of actions are largely controlled by the cortico-basal ganglia-thalamo-cortical loop (Alexander et al., 1986). Movement control is mediated by three distinct pathways in this loop: the direct, the indirect, and the hyperdirect (Rocha et al., 2023). In the direct pathway, dopaminergic neurons originating from the substantia nigra release dopamine in the dorsal striatum, stimulating GABAergic neurons. GABA then binds to GABAergic receptors located in the internal capsule of the globus pallidus. This prevents the globus pallidus from inhibiting the thalamus, allowing the thalamus to innervate the primary motor cortex. Glutamate then activates dopamine D₁-like-containing neurons in the dorsal striatum. The indirect pathway also involves dopaminergic transmission from the substantia nigra to the dorsal striatum. However, dopamine binds to inhibitory dopamine D₂-like receptors. GABA then binds to GABAergic receptors in the external capsule of the globus pallidus, preventing the inhibition of the subthalamic nucleus. Glutamate binds to NMDA receptors in the internal capsule of the globus pallidus, causing GABA release to inhibit the thalamus. As the thalamus sends glutamatergic neurons to motor cortices (Luo et al., 2019), inhibition of the thalamus leads to insufficient stimulation of the primary motor cortex, preventing movement.

Figure 3.

Schematics depicting the glutamatergic pathways involved in motor impulsivity (a), impulsive choice (b), and affective impulsivity (c). The red arrows indicate if glutamate levels are increased or decreased in a region. The dashed arrow indicates that the change in glutamate levels is more speculative as additional research is needed. The curved, dashed black arrows indicate that loss of glutamate homeostasis may occur in a particular brain region. In panel a, the purple arrows indicate GABAergic projections. In panel c, the green arrows represent serotonergic projections. Abbreviations: ACC = anterior cingulate cortex, Amy = amygdala, DRN = dorsal raphe nuclei, GP = globus pallidus, Hip = hippocampus, Hyp = hypothalamus, NAc = nucleus accumbens, OFC = orbitofrontal cortex, PFC = prefrontal cortex, SMA = supplemental motor area, STN = subthalmic nucleus, STR = striatum, TH = thalamus, VTA = ventral tegmental area. Note, the changes in glutamate depicted for motor impulsivity (panel a) are similar to what is observed for affective impulsivity. For simplicity, I have focused on changes in corticolimbic regions in panel c. Created with BioRender.com.

Although dopaminergic neurons originating from the substantia nigra are critical mediators of the direct and the indirect pathways, glutamate-stimulated dopamine release is enhanced in the substantia nigra of SHRs (Warton et al., 2009). Furthermore, motor cortical areas project glutamatergic neurons to the dorsal striatum (Paraskevopoulou et al., 2019), and glutamatergic neurons originating from the cerebral cortex modulate activity of the striatum (Gerfen and Surmeier, 2011), including release of dopamine (Cachope and Cheer, 2014); striatal glutamate also mediates the relationship between dopamine synthesis capacity and striatal activation when individuals need to inhibit a prepotent response (Lorenz et al., 2015). Elevated Glx levels in basal ganglia are observed in children with ADHD and bipolar disorder (Bollmann et al., 2015; Carrey et al., 2007; Castillo et al., 2000; MacMaster et al., 2003; but see Naaijen et al., 2016; Puts et al., 2020), and Glx levels decrease in striatum following medication treatment for ADHD (Carrey et al., 2002; Carrey et al., 2003; but see Carrey et al., 2007). Likewise, SHRs have higher striatal glutamine/glutamate levels and have greater evoked glutamate release in striatum (Miller et al., 2014; Rizzo et al., 2017). Additionally, mice lacking actin depolymerizing factor and n-cofilin, which are present in excitatory synapses, have elevated glutamate release in the striatum and display hyperactive/impulsive behaviors that are attenuated by NMDA receptor blockade and methylphenidate administration (Zimmermann et al., 2015).

As in PFC, the increased glutamine levels observed in striatum of ADHD children disappears during adulthood (Bollmann et al., 2015). Adults with ADHD have decreased glutamate levels in basal ganglia, but lower glutamate levels in basal ganglia correlate with more severe symptoms of inattention as opposed to impulsivity (Maltezos et al., 2014). Striatal glutamate levels are lower in animals that display increased motor impulsivity in a DRL task (Chuang et al., 2021). Interestingly, decreased aspartate aminotransferase and increased glutamine synthetase levels are observed in both PFC and striatum of SHRs relative to WKYs, indicating decreased glutamate synthesis and increased glutamate metabolism, respectively (Dimatelis et al., 2015). Similar to what was discussed for PFC, these results implicate an aberrant x_C⁻ system in striatum, leading to a loss in glutamate homeostasis, which then leads to elevated glutamate levels in striatal regions.

The hyperdirect pathway involves glutamatergic projections from cerebral cortex to subthalamic nucleus. Activity of the internal capsule of the globus pallidus increases, leading to increased inhibition of the thalamus and the primary motor cortex. At first glance, a hyperactive glutamatergic system in the cortico-basal ganglia-thalamo-cortical loop should lead to decreased motor activity, and thus, decreased motor impulsivity. Yet, cerebellar activity also mediates motor control (Jueptner and Weiller, 1998). The deep cerebellar nuclei receive inhibitory inputs from Purkinje cells and excitatory inputs from mossy fiber and climbing fiber pathways (Baumel et al., 2009). The deep cerebellar nuclei innervate the ventrolateral thalamus, which forms a reciprocal loop with motor cortical areas (Ilinky and Kultas-Ilinsky, 2002; Morigaki et al., 2021). While Glx levels in cerebellar cortex are similar in individuals with ADHD and controls (Endres et al., 2015; Rüsch et al., 2010; Soliva et al., 2010; but see Perlov et al., 2010 for increased Glx in left cerebellum of adults with ADHD), children receiving stimulant treatment for ADHD have lower glutamate-glutamine ratios in left cerebellum (Benamor, 2014). This finding is difficult to interpret as most of the participants had the combined subtype (66.67%) or the inattentive subtype (30.39%) of ADHD. Hyperactive cerebellar glutamatergic activity is observed in MDD (Kahl et al., 2020; but see Chen et al., 2014). This hyperactivity may stem from lower GAD-65 and GAD-67 protein levels observed in the cerebellum of individuals with MDD and bipolar disorder (Fatemi et al., 2005; Guidotti et al., 2000). Excess glutamatergic transmission from cerebellum to thalamus would prevent the basal ganglia to adequately inhibit this region, leading to increased movements that interfere with purposeful actions. The result would be increased excitatory outputs from thalamus to primary motor cortex.

There is one potential issue with this interpretation as glutamate/Glx levels are lower in the primary motor cortex of children/adolescents with ADHD (Kahl et al., 2022). One important consideration is that the PFC has reciprocal connections with the premotor cortex (Carmichael and Price, 1995), which itself has reciprocal connections with the primary motor cortex and the supplemental motor area (Ohbayashi, 2021). Hyperpolarization (i.e., reducing activity) of supplemental motor area reduces motor impulsivity in humans (Spieser et al., 2015) whereas impaired inhibitory control is associated with reduced activity of premotor/motor cortex (Weafer et al., 2015). One possible explanation to account for the hypoactive glutamatergic system in premotor/primary cortex is that enhanced glutamatergic transmission from PFC activates inhibitory interneurons in premotor cortex, thus preventing adequate excitation of primary motor cortex. This interpretation is supported by the finding that optogenetic inhibition of premotor inhibitory interneurons increases spontaneous movements in mice (Giordano et al., 2023).

As glutamatergic transmission from primary motor cortex to subthalamic nucleus helps control inhibition of motor actions (Nambu et al., 2002), lower glutamatergic transmission from primary motor cortex may not allow for sufficient stimulation of subthalamic nucleus, resulting in the inability to inhibit competing movements that interfere with the appropriate action (e.g., getting out of a chair during class). This is supported by the finding that use of repetitive transcranial magnetic stimulation to transiently disrupt excitatory neurotransmission in dorsal premotor cortex leads to loss of inhibitory control in healthy participants (Duque et al., 2012). Furthermore, selective reduction of VGLUT in the subthalamic nucleus increases hyperlocomotion, which appears to result from decreased dopamine transporter binding and slower dopamine reuptake (Schweizer et al., 2014). Optogenetic inhibition of subthalamic nucleus increases responses rates during no-go trials in a go/no-go task, indicating increased motor impulsivity; conversely, optogenetic stimulation leads to decreased motor impulsivity (Piszczek et al., 2022).

Beyond the cortico-basal ganglia-thalamo-cortical loop and motor cortices, some limbic structures are implicated in motor impulsivity. The NAc core is connected to the dorsal striatum and controls motor actions related to obtaining reinforcement whereas the shell region mediates the hedonic aspects of reinforcers (Salgado and Kaplitt, 2015). Lesions of the core, but not the shell, increase motor impulsivity (Pothuizen et al., 2005). However, there is evidence that the shell region is involved in motor actions as inactivation of shell selectively increases motor impulsivity (Feja et al., 2014). To further complicate matters, deep brain stimulation of shell increases motor impulsivity whereas stimulation of the core region decreases motor impulsivity (Sesia et al., 2008). These results are difficult to interpret as techniques such as excitotoxic lesions and deep brain stimulation do not isolate the contribution of individual neurotransmitter systems to a behavior. Unfortunately, very few studies have examined the contribution of NAc glutamate to motor impulsivity, but there is evidence that hyperactive glutamatergic activity in NAc mediates motor impulsivity. High impulsive rats, as screened with the 5CSRTT, have reduced GAD-65/67 protein expression; furthermore, selective interference of GAD-65/67 RNA expression in NAc increases motor impulsivity (Caprioli et al., 2014). Tangentially related to this point is that pharmacological disinhibition of NAc via antagonism of GABAergic receptors promotes hyperactivity (Yael et al., 2019).

The hippocampus is another limbic structure implicated in motor impulsivity (Ferland et al., 2014; Mann et al., 2021). There is some evidence that disruption of the glutamatergic system in hippocampus leads to enhanced motor impulsivity. Local genetic deletion of the GluN1 subunit in the CA3 region of the hippocampus increases premature responses in the 5CSRTT while deletion of the GRIA1 gene leads to decreased motor impulsivity (Finlay et al., 2015; Kilonzo et al., 2022). These results indicate that increased AMPA-to-NMDA receptor ratios (i.e., increased LTP resulting from increased excitation) exacerbates motor impulsivity. Increased hippocampal glutamatergic activity could contribute to elevated glutamate levels in NAc as the hippocampus has direct glutamatergic inputs to NAc (e.g., Floresco et al., 2001). The argument for a hyperactive hippocampal glutamatergic system is somewhat challenged by the finding that mice lacking TARP γ-8 exhibit hyperactivity and show reduced AMPA receptor protein expression and AMPA receptor-mediated synaptic transmission in hippocampus (Bai et al., 2022). Downregulation of AMPA receptor protein expression was not observed in PFC of TARP γ-8 knockout mice, but examination of other brain regions was not included. Thus, determining if increased hyperactivity results exclusively from impaired hippocampal AMPA receptor functioning is difficult, especially as motor impulsivity was not tested in conjunction with hyperactivity. As discussed in the next section, a hypoactive hippocampal glutamatergic system is implicated in impulsive choice.

3.5.3. Glutamate and Impulsive Choice.

Impulsive choice is known to be regulated by frontal cortical and limbic regions (Dalley et al., 2011). As discussed in Section 3.5.1., there is evidence that hypoactive intra-PFC glutamatergic activity is associated with impulsive choice. The OFC is another frontal cortical structure that is often studied, but discrepancies are reported in the literature concerning the contribution of OFC to impulsive choice. When whole OFC is lesioned/damaged or pharmacologically inactivated, increases (Berlin et al., 2004; Mobini et al., 2002; Kheramin et al., 2002), decreases (Winstanley et al., 2004), and unaltered (Abela and Chudasama, 2013; Jo et al., 2013; Mariano et al., 2009) impulsive choice are reported. The discrepancies may partially be explained by baseline levels of impulsive choice and the presence of cues during the delay to receiving the large, delayed reinforcer (e.g., Zeeb et al., 2010). More importantly, subregions of OFC have distinct functions, with mOFC representing reward value and lOFC representing punishment and non-reward (see Rolls et al., 2020). However, discrepancies are still observed when a specific region of OFC is targeted. Damage to mOFC increases impulsive choice in humans (Peters and D’Esposito, 2016), but lesions to mOFC decrease impulsive choice in rodents (Mar et al., 2011; but see Stopper et al., 2014 for null effects following pharmacological inactivation). Stronger recruitment of mOFC is associated with increased impulsive choice in individuals with ADHD (Wilbertz et al., 2012), an effect that may be influenced by increased Glx levels in PFC. As the studies cited here did not selectively alter glutamatergic signaling, the contribution of intra-OFC glutamate to impulsive choice needs to be examined. As such, I will limit my discussion to how dysregulation of prefrontal glutamatergic circuits drives impulsive choice (see Fig. 3b).

The PFC exerts top-down control of limbic structures to regulate executive functioning (Del Arco and Mora, 2009). The amygdala has reciprocal connections with both hippocampus (Zhang et al., 2021b) and OFC (Zald and Kim, 1996), and the hippocampus projects neurons back to the OFC (Wikenheiser and Schoenbaum, 2016). As such, corresponding reductions in hippocampal and amygdalar glutamate levels should be associated with impulsive choice. There is evidence to support hypoactive glutamatergic signaling in hippocampus of individuals that display increased impulsive choice. Excitatory neurotransmission is blunted in hippocampus of SHRs (Jensen et al., 2009). Individuals with MDD have decreased hippocampal Glx levels (Block et al., 2009; Zeng et al., 2023; but see Hermens et al., 2015) and decreased VGLUT mRNA expression in hippocampus and surrounding entorhinal cortex and middle temporal gyrus (Medina et al., 2013; Uezato et al., 2009). Even during the euthymic phase (i.e., absence of mood disturbance), increased glutamine and decreased glutamate levels are observed in those with bipolar disorder (Gupta et al., 2022), and decreased VGLUT mRNA expression is observed in entorhinal cortex of individuals with bipolar disorder (Uezato et al., 2009). The lower glutamate levels are not necessarily due to differential EAAT expression as decreased hippocampal expression of EAATs is observed in individuals with MDD, with a similar trend being measured in those with bipolar disorder (Medina et al., 2013). Somewhat similarly, compared to WKYs and Sprague Dawley rats, SHRs have reduced GLT-1 protein expression in hippocampus, but the GLT-1b splice variant is elevated in SHRs in this region (Sterley et al., 2016).

Additional evidence for a hypoactive hippocampal glutamatergic system comes from studies examining ionotropic glutamate receptors. While hypermethylation of the GRIN2A gene is observed in hippocampus of individuals with MDD (Kaut et al., 2015), individuals with bipolar disorder have reduced GluR2, GluR3, and GluR6 mRNA levels in entorhinal cortex (Beneyto et al., 2007), as well as reduced PSD-95 density levels in hippocampus (Toro and Deakin, 2005). SHRs have reduced AMPA receptor-mediated neurotransmission in hippocampus (Medin et al., 2019), but they have increased hippocampal GluN2B expression (Jensen et al., 2009). Furthermore, GluA1 knockout mice show elevated impulsive choice, albeit this effect is modest (Barkus et al., 2012). Although SHRs have higher GluN2B expression, NMDA receptor density is downregulated in both hippocampus and superior temporal cortex in individuals with MDD and bipolar disorder (Nudmamud-Thanoi and Reynolds, 2004; Scarr et al., 2003). Relatedly, reduced hippocampal NMDA receptor functioning is observed in individuals with bipolar disorder (Chitty et al., 2015). Finally, ionotropic glutamate receptor subunit expression (GluR1, GluR3, GluN1, GluN2A, and GluN2B) is downregulated in perirhinal cortex in both MDD and bipolar disorder (Beneyto et al., 2007), a region that has direct connections with the hippocampus (Liu and Bilkey, 1996).

A hypoactive hippocampal glutamatergic system provides an account of the memory disturbances that are commonly observed in individuals with MDD (e.g., Au et al., 2013; Bearden et al., 2006; Cao et al., 2016; Shan et al., 2018; Zaremba et al., 2019) and in SHRs (Grünblatt et al., 2015; Lin et al., 2022). Impaired working memory is linked to impulsive choice (Khurana et al., 2017; Szuhany et al., 2018), and there is some evidence that the hippocampus supports working-memory processes in humans and animals (Borders et al., 2022; Courtney, 2022; Cristoforetti et al., 2022; Hauser et al., 2020; Yoon et al., 2008; but see Slotnick, 2022 for a critical review). Thus, deficient glutamatergic signaling in hippocampus may lead to impaired working memory, which, in turn, can increase impulsive choice in individuals.

While increased decision-making deficits are observed in individuals with amygdalar damage (Gupta et al., 2011), there is not as much evidence supporting hypoactive amygdalar glutamatergic-induced augmentation of impulsive choice. Increased VGLUT protein expression and decreased GAD-67 protein expression are observed in individuals with MDD and bipolar disorder (Varea et al., 2012). These results suggest a hyperactive glutamatergic system, but the increased VGLUT and decreased GAD-67 protein expression may result from deficient glutamate release in the synapse (i.e., compensatory regulation). The reduced glutamate is then unable to effectively regulate amygdalar activity, predisposing at-risk individuals to increased impulsive choice. More recent work using an optogenetic approach to inactivate excitatory neurotransmission during a delay-discounting task reveals complex regulation of impulsive choice by basolateral nucleus of amygdala (BLA). When BLA is inactivated before a choice is made, impulsive choice decreases; however, when this region is inactivated during receipt of a small, immediately delivered reinforcer, impulsive choice increases, but this effect is observed in younger rats only (Hernandez et al., 2019). These results indicate that a hyperactive BLA biases individuals toward immediate reinforcement.

Although there is evidence linking reduced glutamate in hippocampus and, to a lesser extent, amygdala to impulsive choice, enhanced glutamatergic signaling in ACC, which receives inputs from PFC and amygdala (Sarawagi et al., 2021), leads to increased impulsive choice. When glutamate is analyzed in isolation from Glx, adults with ADHD have higher glutamate levels in ACC (Bauer et al., 2018; but see Ende et al., 2016), and women with borderline personality disorder and comorbid ADHD and borderline personality disorder show increased glutamate levels (Hoerst et al., 2010; Rüsch et al., 2010). Glutamate levels are positively correlated with hyperactivity and impulsivity in individuals with ADHD (Bauer et al., 2018). This finding is consistent with the positive correlation between glutamate concentrations in ACC and self-reported impulsive choice (Hoerst et al., 2010), as well as delay discounting in healthy controls (Schmaal et al., 2012a) and in cocaine-dependent individuals (Schmaal et al., 2012b; but see Subramaniam et al., 2022 for null effects with adolescent cannabis users). N-acetylcysteine reduces glutamate levels in the dorsal ACC of cocaine-dependent individuals, with higher levels of BIS-11 impulsivity scores predicting N-acetylcysteine-induced decreases in glutamate (Schmaal et al., 2012b). Presentation of drug-paired stimuli leads to increased intra-ACC glutamate levels (Engeli et al., 2021). These results suggest that increased glutamate in ACC, like PFC, drives impulsive behaviors. The increased ACC glutamatergic activity may occur because hypoactive glutamatergic signaling from PFC is not able to sufficiently stimulate inhibitory parvalbumin interneurons that are highly concentrated in this region (van Heukelum et al., 2019).

Like ACC, increased activity of NAc is associated with impulsive choice in humans (Hamilton et al., 2020; Hariri et al., 2006; Wittmann et al., 2010), and NAc glutamate levels increase following presentation of drug-paired stimuli (Engeli et al., 2021). Paradoxically, SHRs have blunted glutamate-stimulated dopamine release in shell region relative to core region, an effect that is absent in WKYs (Russell, 2003). Inactivating either subregion of NAc increases impulsive choice (Feja et al., 2014), and lesions to the core, but not the shell, increase impulsive choice (Pothuizen et al., 2005). Impulsive choice is also increased in rats with low baseline levels of impulsivity following chemogenetic inhibition of the mPFC-NAc pathway (Wenzel et al., 2023). One potential explanation to reconcile these inconsistent results is that there is a loss of glutamate homeostasis in NAc, which is a central argument of the glutamate homeostasis hypothesis of addiction (Kalivas, 2009). There may be insufficient stimulation of presynaptic mGluRs, resulting in elevated glutamate levels in the NAc that stimulate excitatory receptors (e.g., NDMA). Indeed, Group II and Group III mGluRs are expressed in NAc (Ohishi et al., 1995; Ohishi et al., 1998), and intra-NAc infusion of the GluN2B-selective antagonist ifenprodil, but not the AMPA receptor antagonist CNQX, decreases impulsive choice (Yates and Bardo, 2017). This finding is relevant as GluN2B-containing NDMA receptors are primarily extrasynpatic (Vieira et al., 2020). By blocking GluN2B-containing NMDA receptors, extrasynaptic glutamate can be exchanged for cysteine via the x_c⁻ system and/or stimulate presynaptic mGluRs.

3.5.4. Glutamate and Affective Impulsivity.

Figure 3c shows the glutamatergic pathways implicated in affective impulsivity. The circuitry involved in affective impulsivity is highly similar to the circuitry controlling motor impulsivity. Hyperactivity in PFC leads to overexcitation of motor control centers, contributing to increased motor impulsivity associated with affective impulsivity (e.g., aggression toward oneself or to another individual). Furthermore, the PFC sends glutamatergic neurons to the dorsal raphe nuclei, where serotonergic cell bodies are located (Michelsen et al., 2008). The raphe nuclei contain GABAergic interneurons (Hernández-Vázquez et al., 2019); as such, increased glutamatergic neurotransmission from PFC leads to increased inhibition of serotoninergic neurons (Soiza-Reilly and Commons, 2011). Serotoninergic hypoactivity is highly associated with suicide (Mann et al., 1990) and aggression (Seo et al., 2008). Deficient serotonin release from the dorsal raphe nuclei leads to disinhibition of the hypothalamus, a region where increased activity is linked to suicide (Bao and Swaab, 2018) and aggression (Hashikawa et al., 2017).

In addition to increased hypothalamic activity, limbic structures become hyperactive. Not only does ACC receive glutamatergic inputs from PFC and amygdala, BLA and mPFC have reciprocal glutamatergic projections (Zimmerman et al., 2019), and the hippocampus sends glutamatergic neurons back to the PFC (Zhou et al., 2020). Descendants of individuals that died by suicide have elevated gene expression for mGluR₅ and Homer 1a (García-Gutiérrez et al., 2023). Individuals that died by suicide have increased AMPA receptor (GluR1–4), kainate receptor (subunits 1 and 3), GluN1, GluN2A, GluN2B, mGluR₁, mGluR₂, VGLUT, and PSD mRNA expression (Zhao et al., 2018), and men that died by suicide have decreased EAAT in ACC (Powers et al., 2020). Relatedly, quinolinic acid levels are increased in ACC of individuals that died by suicide (Steiner et al., 2011). Overexcitation of amygdala, hippocampus, and ACC may lead to excitotoxicity given the large number of ionotropic glutamate receptors in these regions (Monaghan and Cotman, 1985), thus accounting for neuropathology (e.g., atrophy) observed in these regions of individuals that have attempted suicide or in individuals/animals that engage in aggressive behavior (Pardini et al., 2014; Parisi et al., 2021; Roberts et al., 2021; van Heukelum et al., 2021; Zetsche et al., 2007; Zhang et al., 2022; but see Schmaal et al., 2020 for a discussion of inconsistent findings regarding amygdala volume and suicide). Although I did not discuss the ACC as related to motor impulsivity, chemogenetic inhibition of ACC pyramidal cells reduces premature responses in the 5CSRTT (van der Veen et al., 2021). This can be related to affective impulsivity as an individual with a condition like MDD or bipolar disorder may be more likely to engage in self-injurious behavior as a result of a hyperactive ACC.

The increased excitation observed in prefrontal cortical areas and the limbic system accounts for increased NAc activity during aggressive behavior (Chester and DeWall, 2016; Dai et al., 2022; Golden et al., 2019) as glutamatergic inputs mediate dopamine release in this region (Chéramy et al., 1998; Imperato et al., 1990). In addition to receiving glutamatergic inputs from PFC, the NAc receives glutamatergic inputs from hippocampus, amygdala, and ACC (Scofield et al., 2016). Increased NAc activity is also observed when individuals with borderline personality disorder, but not healthy controls, are presented with images of painful stimuli while in a state of heightened distress (Olié et al., 2018). Similarly, individuals that engage in non-suicidal self-injury have increased functional connectivity between NAc and lingual gyrus (in occipital cortex) (Chen et al., 2023), and individuals that have recently attempted suicide have higher NAc volume (Kim et al., 2021). A hyperactive NAc can explain why forms of affective impulsivity, particularly self-harm, are negatively reinforcing (Chapman et al., 2006).

4. Discussion

Glutamate is the most abundant neurotransmitter in the mammalian brain. According to Gasiorowska et al. (2021), over 90% of all neurons express glutamate receptors, and approximately 40% of neurons release glutamate as a neurotransmitter. Considering the ubiquity of glutamate in the central nervous system, there should be little surprise that dysfunctional glutamatergic systems are now implicated in the pathogenesis of numerous conditions like Alzheimer’s disease (Wang and Reddy, 2017), schizophrenia (Kruse and Bustillo, 2022), autism spectrum disorder (Nisar et al., 2022), ADHD (Lesch et al., 2013), mood disorders (Jun et al., 2014), and addiction (Kalivas, 2009; Pettorruso et al., 2014). While the current review focuses on glutamatergic aberrations observed in the last three disorders listed above, impulsive behaviors are commonly observed in each of these conditions, especially impulsive choice (Ahn et al., 2011; Chantiluke et al., 2014; Faja and Dawson, 2015; Geng et al., 2020; Heerey et al., 2007; Thoma et al., 2017; Wang et al., 2020a; Warnell et al., 2019; Yu et al., 2017) and affective impulsivity (Banwari et al., 2013; Caspi, 2015; Fritz et al., 2020; Giannouchos et al., 2023; Kaartinen et al., 2014; Leclerc et al., 2018; Radomsky et al., 1999; Ryden et al., 1991; Serafini et al., 2016; Stark et al., 2022).

4.1. Potential Future Directions

In the current review, I propose that motor impulsivity and affective impulsivity can be attributed to a hyperactive glutamatergic system in PFC that causes downstream dysregulation of motor control regions and limbic system while impulsive choice is characterized by a hypoactive PFC and a combination of hypoactivity and hyperactivity in limbic system. Most of the evidence linking hyperactive glutamatergic systems to impulsive behaviors comes from glutamate/Glx measurements in individuals with a psychiatric disorder and from behavioral pharmacology experiments. One challenge associated with using psychiatric populations is dissociating glutamatergic control of impulsive behavior from other symptoms such as depressed mood. Increased use of nonclinical populations in MRS studies can help minimize this caveat. Yet, one potential limitation associated with MRS studies is differentiating glutamatergic dysfunction from a dysregulated GABA system. The glutamatergic and the GABAergic systems are highly intertwined, with GAD converting glutamate into GABA and GABA transaminase catalyzing GABA into glutamate. Thus, elevated glutamate levels observed in individuals may stem from deficient GABA transmission in interneurons in regions like PFC. Behavioral pharmacology experiments can also be limited as ligands that are commonly used to manipulate the glutamatergic system often have other mechanisms of action. For example, some NMDA receptor antagonists have high affinity for serotonergic and adrenergic receptors (Chenard et al., 1991; Gill et al., 2002; Kapur and Seeman, 2002; Rammes et al., 2001). To circumvent these limitations, additional methods can be used to better elucidate how glutamate drives impulsive behaviors.

Instead of using genetic models of ADHD in which a protein/enzyme has been widely suppressed throughout the central nervous system, a method such as small-interfering RNA (siRNA) can be used to suppress protein expression in a specific region. Alternatively, optogenetic/chemogenetic approaches can be used to selectively infect neurons in brain regions known to have a high expression of glutamatergic neurons. Of the studies cited in the current review, only a few used an approach in which glutamatergic neurons were selectively infected (e.g., van der Veen et al., 2021). Given the concerns about the validity of existing animal models of ADHD such as the SHR (Alsop, 2007; Leffa et al., 2019; Sontag et al., 2010), using genetic approaches can help us further our understanding of the glutamatergic basis of impulsivity.

One emerging field in the treatment of disease is pharmacogenomics (Kabbani et al., 2023). Recent evidence shows that glutamatergic genes are important mediators of treatment outcomes in individuals with conditions like ADHD and bipolar disorder. Specifically, children with the G/G genotype of a specific GRM7 SNP (rs37952452) do not respond as well to methylphenidate treatment relative to children with the G/A genotype (Park et al., 2014). However, individuals with the C/G genotype of the rs2284411 SNP of the GRIN2B gene respond better to methylphenidate treatment (Kim et al., 2017). Using an epigenome pathway analysis, Higgins et al. (2015) show that a glutamatergic network, including CACNG2, CACNA1C, GRIA2, and SLC1A2 genes, mediates lithium response in individuals with bipolar disorder. Increased understanding of the genes involved in impulsive behaviors can help therapists design pharmacological interventions that are highly specific to an individual, which can help improve treatment outcomes.

Considering that glia cells are necessary for glutamatergic functioning, these cells may provide novel approaches for ameliorating impulsive behaviors. As noted in Section 2, glutamine is released from neighboring glia cells, and the x_c⁻ system is located on astrocytes and microglia (Bridges et al., 2012). Disrupting microglia development during adolescence attenuates high fat diet-induced increases in motor impulsivity later in life (Smith et al., 2020). Exposure to in utero inflammation increases hyperactivity in mice and increased expression of three receptor subtypes that are localized to microglia (Makinson et al., 2019). Mood-stabilizing drugs like lithium cause a downregulation of kainate type 2 receptors located on astrocytes, with no effect on kainate receptors located on neurons (Li et al., 2009), and long-term valproate treatment upregulates EAAT-1 mRNA levels in cultured glia cells (Aguirre et al., 2008). Collectively, these results suggest that decreasing glutamatergic activity at glia cells can improve aspects of impulsivity.

4.2. Conclusions

Both clinical and preclinical studies implicate aberrant glutamatergic systems in the etiology of impulsivity. Specifically, elevated glutamate from prefrontal cortical areas causes hyperactivity of motor control regions, leading to increased motor impulsivity. A hyperactive PFC also leads to over excitation of the limbic system, accounting for forms of affective impulsivity. The shared glutamatergic perturbations observed for motor impulsivity and affective impulsivity can explain why these forms of impulsivity are often correlated with each other (e.g., Armiya’u et al., 2020; Huang et al., 2023; Liu et al., 2019; Park et al., 2020; Pawliczek et al., 2013; Wang et al., 2017a). In contrast to motor/affective impulsivity, a hypoactive PFC is observed in impulsive choice. The differential glutamatergic mechanisms underlying impulsive choice may account for why this form of impulsivity is inconsistently correlated with other measures of impulsivity (e.g., Liu et al., 2019; Stojek et al., 2014; Yan et al., 2016; Yan et al., 2022).

The glutamatergic system is becoming an important target for treating disorders characterized by impulsive behaviors, specifically mood disorders. While seemingly paradoxical, research suggests that an important approach for ameliorating motor impulsivity and/or affective forms of impulsivity in clinical populations is to increase glutamatergic transmission by administering an ionotropic receptor antagonist (e.g., ketamine/memantine). The corresponding increases in glutamate can bind to extrasynaptic Group II/III mGluRs, thus restoring glutamate homeostasis. Alternatively, targeting the cystine-glutamate antiporter directly with N-acetylcysteine may be a promising pharmacotherapy for attenuating impulsive behaviors. Further research targeting the glutamatergic system is important for developing novel interventions for those at risk for engaging in maladaptive impulsive behaviors.

Highlights.

• Hyperactive glutamate system underlies motor impulsivity and affective impulsivity

• Hypoactive PFC contributes to impulsive choice

• Loss of glutamate homeostasis in PFC may lead to motor/affective impulsivity

• Loss of glutamate homeostasis in nucleus accumbens may mediate impulsive choice

• Glutamate may be a viable target for treating impulse-control disorders