The Limbic System in Co-Occurring Substance Use and Anxiety Disorders: A Narrative Review Using a Dimensional Framework

1. Introduction

Substance use disorders (SUDs) are highly prevalent and costly. In the United States, drug and alcohol use contributed to nearly 11 million premature deaths in 2019 alone [1]. Effective treatment is particularly challenging due to the significant comorbidity between SUD and other mental health conditions. Among the concomitant mental health diagnoses, mood and anxiety disorders have strong reciprocal relationships with SUD [2,3]. For example, drug withdrawal can result in well-characterized autonomic and somatic symptoms classically accompanied by anxiety symptoms [4,5]. Psychological symptoms accompanying withdrawal, like anxiety, may be mistaken as part of withdrawal rather than an underlying separate mood disorder that requires careful treatment [6]. Aversive affective withdrawal symptoms, such as anxiety, can contribute to the escalation of compulsive drug use, maintenance of use, and relapse after periods of abstinence, i.e. negative reinforcement of drug addiction to alleviate negative emotional states [7,8,9].

Three explanatory models about the potential causal impacts of chronic anxiety on substance use (and vice versa) have been proposed. First, having a psychiatric disorder increases susceptibility to addictive behaviors because certain substances may temporarily reduce symptoms of mental conditions and act as a negative reinforcer [10]. Social anxiety disorder has been identified as an example of such a relationship, but for other anxiety disorders, the temporality with SUD development remains unclear [11]. Second, SUDs may promote vulnerability to anxiety disorders due to the consequences of chronic drug use or related withdrawal symptoms [12]. A study using National Epidemiological Survey on Alcohol and Related Conditions (NESARC) data found a higher likelihood of individuals having alcohol use disorder (AUD) as their primary diagnosis among patients with comorbid panic disorder or generalized anxiety disorder [11]. Third, shared genetic and environmental factors (e.g., traumatic or stressful life experiences) may mediate the relationship between SUDs and anxiety rather than reflecting a direct causal association ([13,14]. Importantly, these three possibilities are not mutually exclusive [15]. Neurobiological study of the pathogenesis and evolution of comorbid SUD and anxiety disorders is crucial to resolve these questions, advance treatment, and improve outcomes.

One approach for this purpose is to use the Research Domain Criteria (RDoC), a multi-system framework that maps domains of neurobehavioral functioning to mental disorders. Prior studies have successfully applied the RDoC framework to produce clinically-relevant insights for depressive, personality, and psychotic disorders [16,17,18]. Here we examine SUD-AD comorbidity through this lens, reviewing studies that highlight limbic system neuroadaptations associated with these disorders. For the purposes of this review, we included anxiety disorder diagnoses from multiple iterations of the DSM (i.e., panic disorder, generalized anxiety disorder, social anxiety disorder) and also included search terms that describe state and trait anxiety symptoms.

2. Dimensional Frameworks for Understanding Comorbidity

RDoC provides a transdiagnostic, systematic framework to characterize shared neurobiological processes underlying SUDs and ADs across domains. The RDoC matrix covers six domains: negative valence, positive valence, cognitive, social processes, arousal/regulatory, and sensorimotor systems. The AARDoC further delineates the major underlying domains of functions relevant to SUD: Negative Emotionality (mapping onto the negative valence system), Incentive Salience (mapping onto the positive valence system), and Executive Function (mapping onto the cognitive system). We note that the social processing, arousal, and sensorimotor systems are also relevant for SUD but here we focus on the three relevant domains suggested by [19].

Current neuroscience-based models conceptualize SUD as a three-stage cyclical process that intensifies with continuing drug use. These stages include 1) binge and intoxication, 2) withdrawal and negative affect, and 3) preoccupation and craving [20]. The three stages in the addiction cycle represent three functional psychobiological constructs: incentive salience or pathological habits, reward deficit or stress surfeit, and executive dysfunction. Anxiety, marked by a state of distress and arousal, is relevant to all three stages of SUD. Given significant overlaps between SUDs and anxiety in various domains, especially negative valence and cognitive systems, this psychobiological framework is of particular relevance to the SUD-AD comorbid condition. Core dysfunctions of SUD and AD and their manifestations were investigated in the context of the RDoC matrices for negative valence, positive valence, and cognitive systems for each disorder.

Below we summarize the basic neurobiology underlying each process, followed by a narrative review of the relevant neuroimaging data.

2.1. Positive Valence Systems–Binge/Intoxication

2.1.1. Neurobiology

Individuals with anxiety often exhibit diminished sensitivity in positive valence systems, including the motivation to obtain rewards [21,22,23]. This is evidenced by decreased activation of the striatum to reward feedback [24]. Blunted reward processing to natural rewards may prime people with anxiety disorders to seek out other experiences that increase reward signaling, such as substance use [25]. During the binge and intoxication phase, consuming drugs increases activity related to positive valence systems, such as reward seeking and habit learning [26]. The mesolimbic and the mesocortical dopamine systems are highly involved in this drug-induced reward processing. The mesolimbic system includes neurons that originate in the ventral tegmental area (VTA) and project upwards to the ventral striatum (VS), including the nucleus accumbens. Drug consumption induces rapid ‘bursts’ of dopamine release in the vs. (also known as phasic dopamine release) [27], which activate dopamine D1 receptors and are associated with the rewarding experience of feeling “high” or euphoric [28,29]. The adjacent nigrostriatal dopamine system has been classically associated with motor control, but it also plays a role in reward processing. This pathway originates in the substantia nigra and projects upward to the dorsal striatum (DS). Like in the VS, drug consumption also increases dopamine release in the DS, though to a lesser extent than in the other dopamine systems [28,29].

With long-term drug use, an individual’s reward processing distorts to disproportionately favor drug consumption over other rewards. This reflects in part long term potentiation (LTP), which is one of the mechanisms driving neuroplasticity in the brain triggered by drug-induced signaling [30]. These neuroplastic changes drive incentive salience i.e. heightened or focused attention toward cues linked to rewarding behaviors [31]. Similar to the mesolimbic system, the mesocortical system also originates in the VTA and includes neuronal projections to the medial prefrontal cortex (mPFC). Release of dopamine into the mPFC cues incentive salience, which increases motivation for further drug consumption [20,29]. This is further evidenced by research that has found that as an individual starts to develop drug-cue associations, phasic dopamine bursts also occur in response to drug cues [32].

2.1.2. Neuroimaging Findings: Anxiety → Binge/Intoxication

Stress, particularly during early life, can sensitize reward circuitry and make someone more prone to engaging in substance use [33]. There is neurobiological evidence for this idea; neuroimaging studies find acute intoxication could potentially counteract anxiety-related neural dysregulation. Anxiety is associated with dampened or dysregulated activity in the vs. [34,35], so during intoxication, the supraphysiological bursts of vs. dopamine and heightened vs. activity may overcome these deficits [20,36,37]. As people consume drugs more frequently, drug cues alone can also prompt increased vs. activity [38]. Further, many anxiety disorders are consistently associated with increased activity in fear-related limbic regions including the amygdala [39], particularly in response to relevant anxiety cues [40]. Similar to how acute intoxication can increase vs. activity to potentially counteract dysregulated reward-related activity in people with anxiety, some studies have found that intoxication, particularly with alcohol, can acutely decrease amygdala reactivity [41,42].

2.1.3. Neuroimaging Findings: Binge/Intoxication → Anxiety

While drugs may temporarily relieve feelings of anxiety, chronic self-medication leads to maladaptive neuroplastic changes that decrease the value of drugs. For example, with long-term substance use, stimulant drugs do not induce an increase in dopamine signaling to the same degree as healthy controls [37]. Simultaneously, changes in the nigrostriatal system may support habit formation [43], which may be why drug cues alone are sufficient to induce dopamine increases in dorsal striatum in SUD [44]). Intriguingly, a case study found that damage to the dorsal striatum completely reversed addictive behavior [45]. Over time, a paradoxical brain response develops; dopamine responses to drug consumption decrease while incentive salience increases [20,46]. This mismatch results in increased “wanting” and decreased “liking” [46].

Impaired reward prediction error (RPE) learning underlies the drive to seek rewards in SUDs [47] and can lead to increased negative emotions and feelings of anxiety, as the experience of drug consumption does not match the anticipated response. Disrupted dopamine RPE signal is associated with neuronal plasticity alterations in the striatum, frontal cortex, and amygdala [48]. Behaviorally, prediction errors are also implicated in anxiety disorders as wrong expectations of danger can trigger excessive fear levels and motivate avoidance behaviors [49]. Clinical studies showed that prediction error signaling is disrupted in patients with generalized anxiety disorder (GAD) completing a passive avoidance task and associated their performance with ventromedial prefrontal cortex (vmPFC) and ventral striatum regions. Individuals with social anxiety disorder (SAD) also show similar trends but with elevated dmPFC activation correlating to prediction errors, and reduced dmPFC-ventral pallidum connectivity [50].

2.2. Negative Valence Systems - Withdrawal/Negative Affect

2.2.1. Neurobiology

Withdrawal/negative affect follows the binge/intoxication stage of the addiction cycle and includes core components of anxiety disorders. During withdrawal, dopamine transmission decreases in mesolimbic circuits are associated with motivational deficits and decreased mood [30]. As dependence and withdrawal develops, the brain’s antireward system is activated and negative affective symptoms emerge including anxiety, irritability, and hyperkatifeia [51]. While different drugs can elicit different withdrawal symptoms, negative affect is a universal and motivational element of withdrawal [52]. These negative emotional states spur further drug use to avoid discomfort, i.e. negative reinforcement [53]. Individuals with a history of anxiety experience more intense withdrawal symptoms and withdrawal-related discomfort and relapse from substances like methamphetamine, nicotine, or alcohol [54,55]), and baseline diagnosis of anxiety disorders like SAD and panic disorder significantly predicted early alcohol relapse [56].

These negative emotional states are driven by increased corticotropin-releasing factor (CRF), noradrenaline, and dynorphin signaling [51]. Injection of CRF antagonists into the central amygdala reduces anxiogenic-like states in rats withdrawn from substances including alcohol, nicotine, and cocaine [57,58,59,60]. Further, noradrenergic antagonists administered into the BNST reversed opioid withdrawal-induced place aversion in rodents [61]. Recent evidence suggests changes in endogenous endocannabinoid systems in the extended amygdala are associated with heightened anxiogenic responses in rats with cannabis withdrawal undergoing classical anxiety paradigms like the elevated plus-maze and the light/dark box [62,63,64]. Clinically, classical medications used to treat alcohol and heroin withdrawal include α-adrenergic drugs that inhibit noradrenergic release. Together, these findings highlight the importance of stress signaling in the extended amygdala, a component of the limbic system that is linked with anxiety disorders as demonstrated in human imaging studies.

2.2.2. Neuroimaging Findings: Withdrawal/Negative Affect ⟷ Anxiety

While the directionality between anxiety and SUD specific to the withdrawal stage remains hard to disentangle through neuroimaging research, we present general findings regarding neurobiological overlap between withdrawal and anxiety phenotypes.

The amygdala is heavily implicated in the withdrawal/negative affect stage and increased connectivity in amygdala-related networks is associated with increased emotional reactivity [65,66]. Alterations in connectivity between large-scale brain networks, such as the salience network (SN, including the insula, anterior cingulate, amygdala and hypothalamus) and the default mode network (DMN, including medial prefrontal and parietal cortex) are critically involved in both withdrawal and anxiety. SN-DMN connectivity is elevated during withdrawal from substances such as tobacco and alcohol [67,68]. Elevated SN-DMN connectivity is also commonly observed in adolescents and adults with anxiety disorders compared to healthy controls [69,70,71,72], and positively correlates with anxiety severity among both anxiety patients and healthy controls [69,73,74].

These connectivity changes are thought to direct attention towards withdrawal-induced physiological sensations, heightening internalizing and anxiety symptoms [75]. For example, both nicotine and alcohol withdrawal [76], self-reported anxiety [77], and irritability [78] have all been linked to elevated amygdala-insula functional connectivity. This may explain why individuals with a history of anxiety experience more intense withdrawal symptoms and withdrawal-related discomfort and relapse from substances like methamphetamine, nicotine, or alcohol [55]), as increased amygdala-insula reactivity has been observed in anxiety patients across all ages [79,80,81,82].

As mentioned above, α-adrenergic drugs are effective in treating alcohol withdrawal symptoms, and imaging evidence supports this. Prazosin treatment improves dysregulation in DMN regions including medial prefrontal cortex (mPFC), which has been associated with severity of alcohol withdrawal symptoms [83] and implicated in anxiety psychopathology [84]. While the SN is involved in processing internal or external stimuli, and mediates the switching between DMN and executive control network (ECN), the DMN plays a central role in self-referential processes (evaluating salience or internal and external cues, remembering the past, and planning the future) implicated in affective disorders like anxiety [85] and in SUD [29]. Dysfunctions in within-network and between-network DMN connectivity is heavily implicated across all anxiety disorders [86], and recent studies have suggested DMN aberrations may underlie anxiety symptoms such as perseverative thoughts in generalized anxiety disorder (GAD) [87,88] and possibly predict anxiety disorders such as GAD and social anxiety disorder [88,89]. Therefore, among the existing resting state networks, the DMN may be of particular interest in comorbid SUD-anxiety conditions during the withdrawal stage.

Chronic drug use has opposing effects on DMN engagements, which may contribute to the predominance of anti-reward systems during the withdrawal/negative affect stage [86]. The DMN can be further split into anterior and posterior portions, and reduced anterior-posterior DMN connectivity has been associated with impaired self-awareness [90], and in addicted persons this contribute to uncontrolled drug-taking [91,92]. Increased posterior DMN involvement may underlie ruminatory behaviors, resulting in distress and negative emotional states like anxiety [93,94], and resting-state studies in individuals addicted to substances like alcohol, heroin, or cannabis found that chronic drug use enhances posterior DMN, and exacerbates the imbalance of anterior-posterior DMN [29,95,96]. Decreased connectivity between the posterior and anterior DMN may contribute to the inability to inhibit negative emotional states, interfering with abstinence and flexible decision-making which could further exacerbate the anxiety symptoms experienced during withdrawal [97].

Emerging evidence also highlights the importance of the bed nucleus of the stria terminalis (BNST) in withdrawal related-anxiety and relapse as a critical node in both the stress and reward circuitries [2]. During withdrawal, the amygdala releases CRF on receptors in the bed nucleus stria terminalis (BNST) [98], and heightened amygdala-BNST connectivity was observed in people undergoing alcohol withdrawal [68,99]. The BNST is involved in withdrawal related-anxiety and relapse, serving as a critical node in both stress and addiction circuitry with connections to multiple limbic and brainstem regions including the amygdala and VTA. fMRI studies using shock paradigms described increased BNST engagement mediating sustained fear responses, and activity correlated with physiological skin conductance and self-reported anxiety ratings [100,101,102,103].

2.3. Cognitive Systems - Preoccupation/Anticipation

2.3.1. Neurobiology

Commonly known as “craving”, the preoccupation/anticipation stage of addiction is characterized by a disruption of executive function and involves the prefrontal cortex. Studies have linked poor executive function (i.e., inhibition, working memory) with early initiation of alcohol, tobacco, and other substance use [104,105,106]. In turn, acute craving decreases the ability to control attention away from drug cues [107,108]. Comorbid anxiety disorder exacerbates drug craving [109] and both cue- and stress-induced craving and anxiety are linked to early attrition from alcohol use disorder treatment; these observations are in part driven by dysregulated hypothalamic-pituitary-adrenal (HPA) axis function [110].

Prior studies found working memory, attentional control, inhibitory control to be among the most common cognitive domains impacted by anxiety as well as SUD. Reciprocal glutamatergic connections between the PFC and the dorsal striatum are heavily involved in the flexible control of these behaviors [111]. Preclinical evidence showed that rats withdrawn from substances like cocaine or methamphetamine were unable to develop long-term potentiation and long-term depression in the nucleus accumbens region after stimulating the PFC, suggesting that chronic substance use depresses PFC-accumbens synapses [112,113]. However, N-acetylcysteine (NAC) treatment prevents relapse by activating cystine-glutamate exchange in the cortico-accumbens circuitry [114], a finding that is consistent in human studies that reported decreased desire to use drugs in response to cocaine cue-reactivity [115].

Persistent prefrontal dopamine D1 receptor signaling is crucial for working memory, a foundational cognitive function [116,117,118]. Nonhuman primate studies showed blocking D1Rs in the PFC impaired learning of novel associations and decreased cognitive flexibility during a working memory task [117,119], and human studies reported that systemic administration of mixed D1/D2 agonist facilitated working memory while D2 agonist had no effect [120]. Striatal dopamine D1 versus D2 receptor signaling is also implicated in flexible control of behavior. Animal studies using the probabilistic reversal learning paradigm found that positive feedback learning is modulated by D1R signaling in the ventral striatum while D2R signaling modulates negative feedback learning. Stimulation of D2R in the ventral or dorsolateral striatum promoted explorative choice behavior [121] and reversal learning was impaired by D2R antagonism but not D1R antagonism in the dorsal striatum [122]. Studies have provided evidence that dorsal striatum D1 medium spiny neurons (MSNs) mediate reward/reinforcement and D2-MSNs mediate aversion [123]. Further, rats completing an instrumental task showed that D1R inhibition and D2 activation both promote the expression of flexible responding after development of habitual or compulsive-like behaviors [124].

2.3.2. Neuroimaging Findings: Anxiety → Preoccupation/Anticipation

During the preoccupation and anticipation phase of the addiction cycle, attention to drug cues is heightened [125,126]. Neuroimaging studies of cue reactivity have found increased activation related to craving for alcohol, nicotine, or cannabis in cognitive control regions, such as the anterior cingulate cortex (ACC) and ventromedial prefrontal cortex (vmPFC) and limbic regions like the ventral striatum [126,127,128,129]. Alcohol cue-elicited activation of the ventral striatum correlated with behavioral measures [128] and nicotine craving correlated with left vmPFC and amygdala activation when viewing smoking-related pictures [126] People with anxiety disorders also exhibit problems with cognitive flexibility [130], and high trait anxiety is linked to low PFC activity during attentional control [131] and dysregulated connectivity between PFC and subcortical regions including the amygdala, basolateral amygdala (BLA), insula, and hippocampus [130,132,133,134,135,136]. In task-based fMRI studies with the stop signal task, amygdala activation positively correlated with trait anxiety, and with the vmPFC during risk taking. Using the same SST task in cocaine-dependent patients [137], vmPFC activation correlated negatively with improved inhibitory control during methylphenidate treatment [138]. Together, these findings suggest that anxiety-related dysregulation in cognitive systems, combined with increased amygdala activation and stress symptoms associated with the preoccupation phase [20], may worsen or reinforce altered cognitive control associated with cue reactivity.

2.3.3. Neuroimaging Findings: Preoccupation/Anticipation → Anxiety

Long-term substance use creates a paradoxical effect in which reward responses are blunted while anticipation heightens [20,46]. During this preoccupation and anticipation phase, many factors promote craving. Neural biomarkers like the insula, hippocampus, and prefrontal cortex are implicated in the craving stage [125]. The insula, particularly its anterior regions, are reciprocally connected to several limbic regions and is responsible for an interoceptive function (integrating autonomic/visceral information with emotion, motivation, and conscious awareness) [139].

SUDs impair interoception [140] and damaged interoceptive awareness is likely one of the contributing factors to why people with SUD continue substance use [141]. Chronic drug use can lead an individual to associate interoceptive cues with the rewards of drugs to trigger cravings, as the body is reminded of the positive effects of drugs and primes the urge to seek drugs. The imbalance between interoception/exteroception causes an increased focus on external stimuli (i.e. drug cues) and a decreased ability to perceive internal bodily states [78,142]. Neuroimaging studies demonstrated insula activity in response to drug cues in patients addicted to cigarettes, cocaine, and alcohol, and correlated insula activity with self-reported cravings [139,143,144,145,146]. SUD patients with lesions to the insula (or even other regions functionally connected to the insula) can remarkably have a complete remission of their addiction and no longer experience drug cravings or relapse [147,148]. The anterior insula is reciprocally connected to several limbic regions, and cue-reactivity fMRI studies on alcohol-related craving found increased connectivity between the left insula-left dorsomedial PFC but decreased connectivity in a network including the ACC, insula, and hippocampus [149].

The insula, particularly the anterior insula, is also heavily implicated in the pathophysiology of anxiety and anxiety disorders [150,151,152]. Alvarez et al. found that individuals with greater anxiety proneness and less perceived control showed greater activity in the anterior insula during anticipation of unpredictable threat [153]. Individuals who are prone to anxiety experience augmented signaling between observed and expected body state, and may use worrying about possible aversive outcomes as a cognitive avoidance response to stressful, unpredictable life events [151,154]. Similarly, addicted individuals with diminished interoceptive abilities may have difficulties regulating internal/external states in response to stressful situations, and turn to further substance use like alcohol as a coping mechanism [155,156,157]. Given the overlap in dysregulated insular activity and interoceptive abilities underlying both SUDs and anxiety disorders, it is likely that anxiety symptoms are amplified or anxiety disorders emerge during the preoccupation/anticipation stage.

3. Conclusions and Future Directions

The neurobiological circuits underlying SUDs and anxiety disorders interact to exacerbate symptoms as individuals cycle through the three stages of addiction. Anxiety can prompt people to consume drugs, which temporarily recruits positive valence systems and increases incentive salience over time. During withdrawal, activity in negative valence systems heightens, and anxiety symptoms increase rapidly. As people enter into the preoccupation and anticipation phase, changes to cue reactivity and interoception are reflected in cognitive systems. The cycle repeats as people become intoxicated again due to lack of cognitive control and increased incentive salience and cue reactivity. As tolerance increases, people binge with increasing drug quantities and neuroplastic changes are reinforced. Each stage of the cycle can worsen preexisting or SUD-induced anxiety. Common findings from task and resting state fMRI are related to these processes as summarized in Figure 1.

Encouragingly, longitudinal studies of people with SUDs in recovery have found that drug-related functional deficits can be at least partially reversed with abstinence. For example, an investigation of people in OUD recovery found improvements in nucleus accumbens function in response to reward [159]. Increased DA transporters after abstinence was observed in methamphetamine users [160], and DA transporter binding in alcohols matched healthy controls after a 4-week period of abstinence [161]. There is also some promising research that anxiety treatment can cause positive functional changes; studies have found decreases in overactivation of regions of the limbic system, such as the ACC [162], amygdala [163], and hippocampus [164]. Decreased activity in these regions correlated with improvement in PTSD [165] and social anxiety [163] symptom severity.

Further research is needed on treatments for comorbid anxiety and SUDs, including studies with a focus on assessing functional changes in the brain. Longitudinal investigations may prove especially useful, as they would provide information about acute versus long-term changes in these disorders. The direction of causality between anxiety and SUDs could also be better understood; it is important to explore whether "self-medication" of anxiety using substances is more prevalent than SUD-inducing anxiety disorders and whether the order of onset reflects differences in underlying neurobiology. Furthermore, future research should focus on sex differences to investigate how sex influences comorbid anxiety and SUD, and to what extent differences are genetically determined or influenced by environmental factors. Variables such as the duration of drug use, abstinence periods, relapse rates, and overdose rates should also be included to better understand the progression and outcomes of SUD. These research aims would help develop more effective treatment and prevention strategies for SUDs and anxiety disorders.