Summary and discussion

Cocaine use disorder is a chronic relapsing disorder characterized by the compulsion to seek and take the drug, loss of control in limiting intake and emergence of a negative state when access to the drug is prevented1. Despite decades of experimental and clinical research, treatment success in cocaine use disorder is poor; there are currently no registered pharmacological treatments11 and although cognitive behavioral therapy is fairly successful9, there is considerable room for improvement. A better understanding of the (neural) mechanisms involved in cocaine abuse and dependence, may lead to the identification of novel treatment targets.

Neuroimaging studies have made a significant effort to identify biomarkers of substance use disorder, which refers to a measurable indicator of (ab)normal biological processes or treatment response326. Previous research primarily focused on reward and cognitive control functions and thus the involvement of the frontostriatal network in the etiology of substance use disorder. Those studies have consistently demonstrated hyperactivity within the ventral striatum (VS) and prefrontal cortex (PFC) during the processing of drug related information, and hypo-activation of the PFC during executive functioning326,327. While the frontostriatal network is strongly connected to the amygdala117, it has only recently been suggested that PFC-amygdala-VS dysfunction may play a key role in habitual drug seeking and deficient emotion regulation16,33, and therefore in the etiology of substance dependence. The findings described in this dissertation provide relevant support of altered PFC-amygdala-VS function and structure in regular cocaine users, which may provide us with a novel biomarker and treatment target of substance use disorder. The findings described in chapter 2-5 provide novel evidence for hyperactivation of the amygdala and its relation to frontrostriatal dysfunction in regular cocaine users, and how this may be modulated by childhood adversity and state anxiety. In addition, the other chapter provide additional evidence on how deficient PFC structure is related to trait impulsivity (chapter 6 and 7) and polysubstance use (chapter 8). In the end of the discussion I propose a model on how amygdala and frontostriatal function is modulated by childhood adversity, state anxiety, impulsivity and polysubstance use, and how this may be related to compulsive behaviour and impaired emotion regulation in regular cocaine users.

Amygdala hyperactivity as a biomarker of cocaine use disorder
Impaired dorsal anterior cingulate cortex (dACC) function in addicted individuals has consistently been associated with impaired inhibitory control18. Because of its connection with the amygdala, the dACC has also been implicated in emotion regulation78. There has been limited research however on whether amygdala function is also impaired in cocaine dependence. In Chapter 2 we demonstrated that regular cocaine users show hyper-responsiveness of the amygdala and reduced functional connectivity between the amygdala and the dACC in response to angry and fearful facial expressions. In Chapter 3 we furthermore demonstrated that the functional connectivity between the amygdala and dACC is reduced in response to visual cocaine-related cues. These results suggest reduced cognitive control in regular cocaine users, in the presence of cocaine-related or emotional stimuli, thereby increasing the risk of compulsive cocaine use.

Childhood adversities and current negative emotional states have long been suggested to play a role in the onset and continuation of substance dependence21,328. Animal studies have already demonstrated that early life stress induces long-term changes within the amygdala and the frontostriatal network35,37,107,111,112,114, whereas acute fear or stress induces dopamine activation with the VS143. Very few studies, however, have tried to translate these findings from animal studies to humans. In Chapter 3 we demonstrated that cocaine users with a history of childhood adversity show reduced functional connectivity between the amygdala and the VS during cocaine cue-exposure. During the same task, cocaine users with high levels of self-reported state anxiety show reduced amygdala-dmPFC coupling. On a behavioral level amygdala-VS coupling has been suggested to play a role in reward learning, the performance of goal-directed actions and risk aversion127-129, whereas amygdala-dmPFC coupling has been suggested to play an important role in emotion regulation133,134. These findings suggest that cocaine users with a history of childhood adversity may be more prone to the development of compulsive and risky behaviour, whereas highly anxious cocaine users may be more prone to impaired emotion regulation during cocaine cue-exposure and thus relapse.

Compulsive drug seeking behaviour despite negative consequences might be the result of an inability to learn to associate certain stimuli with a negative outcomes, due to insufficient activation of the amygdala. However, excessive amygdala activation in response to negative stimuli, as we demonstrated in cocaine users in chapter 2, is more likely to be related to enhanced association of certain stimuli with negative consequences329,330. In chapter 4 we however demonstrated that cocaine users do not differ from non-drug using controls in the ability to learn an association between a stimulus and a negative outcome, using a classical fear conditioning paradigm. On a neural level however, cocaine users however displayed hyper-responsiveness of the amygdala and insula during fear learning. These results suggest that cocaine users are actually hyper-responsive to stimuli that predict a negative outcome, instead of being unable to learn the relation between an event and a negative outcome. Because of the connection between the amygdala and VS, increased responsiveness of the amygdala to stimuli that predict a negative outcome in regular cocaine users, may represent increased negative reinforcement, stress induced relief craving and subsequent continuation of relapse into drug use1,21,156,161.

It remains to be investigated, however, how amygdala hyperactivity relates to habit formation, reward motivation and compulsive behavior despite negative consequences. An important step in understanding this, is to get more insight in how negative outcomes influence neural reward signaling in general. Together with Guido van Wingen, I therefore developed a novel fMRI paradigm, that we applied in a population of healthy, non-drug using males (chapter 5). In this study we demonstrated that the anticipation of a negative outcome reduces the reinstatement of conditioned reward signaling within the VS. Interestingly, a reduction in reward reinstatement was associated with the strengthening of VS-hippocampus connectivity and the strengthening of VS-insula connectivity. With other words, strong functional connectivity between the VS and hippocampus or insula may be essential to reduce reward motivation when the reward is paired with a negative outcome. While the insula may provide body-relevant information to the degree of aversion of a certain stimulus156,182, the hippocampus is involved in the retrieval of (aversive) memory183. Thus, the anticipation of a negative outcome may reduce reward reinstatement by enhancing the retrieval of aversive memory and disrupting the reconsolidation of reward memory. Since cocaine dependence is characterized by a continued motivation for (drug) rewards despite negative consequences, it can be hypothesized that impaired coupling between the VS and hippocampus or insula underlies the compulsive character of substance dependence.

The association between grey matter and trait impulsivity in regular cocaine users
It has often been demonstrated that highly impulsive individuals are at particular risk to transit from recreational to compulsive cocaine use31,44-46. However, while high impulsiveness is strongly associated with cocaine use disorder331, the neural deficits that underlie the relation between trait impulsivity and cocaine use disorder have scarcely been investigated. Compared to clinical ratings however, these ‘neurocognitive endophenotypes‘ may provide us with a more accurate measure to predict or improve treatment outcome326,332. In chapter 6 we found smaller grey matter (GM) volume of the middle frontal cortex of cocaine users compared to non-drug using controls. More importantly, we demonstrated in cocaine users that trait impulsivity was strongly associated with alterations in GM volume of the orbital frontal cortex, the precentral gyrus, superior frontal gyrus and inferior parietal gyrus.

Because this study aimed to explore the potential confounding effects of impulsivity, depression and smoking on cortical volume abnormalities in cocaine users, we did not test the relation between trait impulsivity and cortical volume in non-drug using controls. Based on the results of other studies however, it became apparent that the association between cortical structure and impulsivity may be different in cocaine users and non-using controls44,50-52,236, although this had only been tested directly once53. Therefore we conducted another structural MRI study to investigate whether cocaine users and controls differed with regard to the association between cortical structure and trait impulsivity (chapter 7). The methods used in this study deviated from the study described in chapter 6 in several ways. First, we included a larger, although partly overlapping, sample of cocaine users and non-drug using controls and second, we applied a surface-based analysis instead of a voxel-based analysis of cortical morphometric, that allowed us to measure cortical thickness and cortical surface area in addition to cortical volume. As expected, we demonstrated a significant group by impulsivity interaction on cortical surface area of the superior temporal cortex and insula and cortical thickness of the dACC. This distinct relation between cortical structure and trait impulsivity could underlie a functional imbalance between the insula, superior temporal cortex and dACC, which has previously been related to impaired decision making75,263-267,269,295,333. The results of this study suggest that a distinct relation between trait impulsivity and cortical structure, in regular cocaine users compared to non-drug using controls, may be an important characteristic of cocaine use disorder334. While it remains to be investigated whether these alterations are a cause or a consequence of cocaine use, these findings may have relevant treatment implications: For instance, it has previously been demonstrated that modafinil (which is a cognitive enhancer suggested to improve PFC function335,336) is only effective in high-impulsive alcohol dependent patients and may even have a detrimental effect in low-impulsive alcohol dependent patients337. A distinct relation between trait impulsivity and underlying cortical pathology may contribute to the prediction of these differential treatment outcomes.

Cocaine users do not only use cocaine
None of the cocaine users that were included in the studies described in this dissertation were using only cocaine: 81% were also cigarette smokers, 47% were also heavy alcohol users (more than 21 units of alcohol per week) and 39% also used cannabis on a (more than) weekly basis. While these numbers of poly-substance use are in line with what is commonly reported in cocaine users57,249 it does raise several questions. Most importantly, are abnormalities in the amygdala and frontostriatal circuitry, as described above in cocaine users, related to cocaine abuse, or rather to tobacco, alcohol or cannabis (ab)use, or a combination thereof? To answer this question, we tested for potential confounding effects of drugs other than cocaine in most of the studies. Comorbid use of nicotine, alcohol and cannabis, was not significantly associated with hyper-activation of the amygdala in response to angry or fearful faces (chapter 2). We also did not find any relation between nicotine or alcohol use and neural responsiveness during fear conditioning or extinction learning (chapter 4). Furthermore, in chapter 6 and 7, we found that nicotine and alcohol use was also unrelated to cortical volume, thickness or surface area. However, cannabis use was associated with reduced activation of the superior temporal cortex during fear extinction learning, although it did not explain neural differences between cocaine users and non-drug using controls during fear conditioning or extinction learning (chapter 4). In addition, cannabis use significantly modulated the group by impulsivity interaction on cortical surface area of the temporal cortex (chapter 7). Thus, while our studies consistently demonstrated that alcohol and nicotine use are unrelated to functional or structural neural abnormalities in cocaine users, there are some indications that cannabis use may normalize or prevent structural and functional alterations in the temporal cortex related to cocaine abuse and dependence (chapter 4 and ). These findings are in line with several studies in alcohol use disorder that demonstrated more severe structural alterations in grey and white matter of heavy alcohol users, compared to heavy alcohol users that also use cannabis273,290,291.

Unfortunately, the relation between polysubstance use and structural alterations in the brain of cocaine users has never been extensively studied. In chapter 8 we therefore performed a large diffusion-tensor imaging study in 67 cocaine users and 67 non-drug using controls. Study participants were classified in subgroups based on the number of substances used (ranging from nothing, to the combined use of cocaine, heavy alcohol and cannabis). In contrast to previous studies, cocaine users that also used cannabis did not show less sever alterations in white matter structure, compared to cocaine users that did not use cannabis. On the contrary, we demonstrated a negative relation between the number of substances used and white matter integrity: participants that reported no use of alcohol or drugs whatsoever, showed the least white matter alterations, whereas polysubstance users of cocaine, alcohol and cannabis showed the largest white matter alterations. While this effect was detected in the white matter throughout the whole brain, the strongest effects were detected within the frontal lobe. As various regions within the frontal lobe play an important role in drug-craving, impulsive and compulsive behavior and impaired decision making230, this could explain why treatment outcome is poorer in poly-substance users compared to single-substance users9,57. An important implication of these findings is that polysubstance users may benefit most from treatment strategies that target PFC dysfunction.

Are neural alterations a cause of cocaine abuse, a consequence of cocaine abuse, or a combination of both?
An important limitation of cross-sectional studies as the ones described in this dissertation, is that they not allow to differentiate between neural differences that predispose to cocaine abuse and those that are the result of cocaine use. Our findings do however allow us to speculate on this matter. For instance, in chapter 2 we demonstrated that cocaine users with a relative early onset of cocaine use or a longer duration of cocaine use, show higher amygdala activation in response to negative emotional facial expressions, compared to cocaine users with a relative late onset or short duration of cocaine use. An early exposure to cocaine may have interfered with the neural development of the amygdala, which would support a causal relationship between cocaine exposure and amygdala hyperactivity. Alternatively, (environmental) stress related to an early onset of cocaine use may also have affected brain development. In support of this, we demonstrated in chapter 3 that functional coupling between the amygdala and VS during cocaine cue exposure, is strongly reduced in cocaine users that experienced high levels of childhood adversities. These findings indeed suggest that stressful events during childhood or adolescence may induce changes within the amygdala-VS network and thereby could predispose to cocaine abuse and dependence.

Other findings in this dissertation also argue against the hypothesis that the neural alterations found in regular cocaine users are simply the consequence of cocaine exposure. For example, there was no association between the amount of cocaine use and the neural response to fear conditioned stimuli (chapter 4). There were also no significant associations between the duration (or onset) of cocaine use on gray matter abnormalities (chapter 6 and 7) or between the amount of substance use and white matter abnormalities (chapter 8). Altogether, these studies do not provide a clear-cut answer on the causal relation between abnormalities in the PFCÐamygdala-striatal circuitry and cocaine use disorder. Nonetheless, the lack of associations between onset age, years and amount of cocaine use in most of our studies, suggest that these alterations may be predominantly predisposing for the development of cocaine dependence rather than just the consequence of chronic cocaine use.

Summary and future perspectives
While previous research primarily focused on the involvement of the frontostriatal network in the etiology of substance use disorder, it has recently been suggested that the amygdala, and its interaction with the frontostriatal circuitry, may play a key role in habitual drug seeking16,33, and therefore in the etiology of substance use disorder. The studies described in this dissertation not only support the involvement of the PFC-amygdala-striatal circuitry in cocaine abuse and dependence, but they also provide novel and more specific evidence for the involvement of this network in cocaine use disorder.
We demonstrated that cocaine users show hyper-activity of the amygdala in response to negative emotional facial expression, cocaine-related stimuli and fear-conditioned stimuli. We furthermore demonstrated that childhood adversity disrupts amygdala-VS coupling during cocaine cue-exposure. As amygdala-VS coupling plays a key role in goal-directed learning127-129 these findings suggest that a history of childhood adversity may predispose to substance abuse by increasing habit learning and reducing risk aversion. We also showed that state anxiety disrupts amygdala-dmPFC coupling during cocaine cue-exposure. In addition, we demonstrated that trait impulsivity and the degree of polysubstance use selectively alters (dm)PFC structure. Because the amygdala-dmPFC coupling plays a key role in emotion regulation133,134, these results imply that negative emotional states, trait impulsivity and polysubstance use may reduce the ability to regulate emotional responses to cocaine related cues, and thus increases the risk of continued of drug use and relapse.
Finally, in healthy, non-drug using individuals we demonstrated that anticipation of a negative outcome reduces reward reinstatement in the VS, in which a strong VS-hippocampus coupling seems to be essential. Since cocaine dependence is associated with a persistent motivation for reward despite negative consequences, it could be hypothesized that negative outcomes do not reduce reward reinstatement within the VS, possibly because of a weaker coupling between the VS and hippocampus. The hippocampus is closely connected to the amygdala via which it can influence amygdala responses when emotional stimuli are encountered338. Although we did not directly test neural responsiveness during the anticipation of a stimulus with both a positive (rewarding) and negative outcome in cocaine users, it has consistently been demonstrated that drug-cue exposure enhances negative emotional states in substance dependent patients e.g.339-341. It can therefore be assumed that cocaine cue-exposure not only triggers (drug) reward anticipation but also the anticipation of a negative outcome. Our finding of reduced amygdala-VS coupling during cocaine cue-exposure in cocaine users with a history of childhood adversity, therefore may imply that negative consequences associated with drug use do not lead to a reduction of reward motivation in these individuals because of reduced striatal coupling with the hippocampus through its connection with the amygdala.
Based on the findings described in this dissertation are summarized in a schematic model, in figure 1: Cocaine users have reduced amygdala-VS coupling, which is thought to be related to compulsive behavior. These alterations may predispose to the development of a cocaine use disorder, possibly linked to childhood adversity. Additionally, cocaine users have reduced amygdala-dmPFC coupling, which is suggested to be related to impaired emotion regulation. These alterations may therefore contribute to the negative reinforcement mechanisms that drive cocaine use. Amygdala-dmPFC coupling is likely to be most impaired in high impulsive individuals and polysubstance users, because of structural PFC abnormalities. The validity of this model, however, remains to be investigated as several relevant questions remain to be unanswered: For instance, how do abnormalities within the amygdala-frontostriatal circuitry relate to an overreliance on habit-behavior? Are these abnormalities specific to stimulant use disorder, or can these results be generalized to other substance use disorders or behavioral addictions. And are these abnormalities truly predisposing or do the abnormalities normalize after a certain period of abstinence. Can alterations within the amygdala-frontostriatal circuitry predict treatment outcome? Cross-diagnostic studies (including for instance cocaine dependent patients and alcohol dependent patients), family studies and longitudinal studies in addition to the use of relevant behavioral paradigms, would be very helpful in answering these questions.



Figure 1. Model on the involvement of the dmPFC-amygdala-VS circuitry in cocaine use disorder.
This model, partluy based on a model proposed by Li and Sinha in 200833, shows how dmPFC-amygdala-VS alterations may be involved in cocaine use disorder. While reduced amygdala-VS coupling is suggested to be involved in compulsive drug seeking, reduced amygdala-dmPFC coupling is suggested to be involved in impaired emotion regulation. Reduced amygdala-VS coupling may be related to childhood adversity, whereas reduced amygdala-dmPFC coupling may be related to state anxiety, trait impulsivity and polysubstance use.
Although the validity of this model remains to be tested, it may provide us with novel treatment targets. For instance, there is increasing evidence for the effectiveness of repetitive transcranial magnetic stimulation (rTMS) in the treatment of substance use disorders, when stimulating the right dorsolateral PFC (DLPFC)342,343. While this approach is thought to be effective mainly due to enhancing the prefrontal control over the striatum342,344, studies in depressive populations not only demonstrated that DLPFC stimulation modulates amygdala function, they also demonstrated that the effectiveness of DLPFC stimulation may dependent on baseline amygdala function345. In patients with obsessive compulsive disorder it has been demonstrated that DLPFC stimulation using rTMS significantly modulates frontal-amygdala connectivity135. Based on these studies, combined with the studies described in this dissertation, it could therefore be expected that rTMS as a treatment of substance use disorder is most effective in individuals that show the strongest abnormalities in amygdala-VS coupling, which are likely the ones that report high levels of state anxiety and childhood adversities. As we also demonstrated that PFC function is most affected in highly impulsive individuals and poly-substance users, this may further suggest that rTMS may also be more effective in cocaine users that are highly impulsive and in cocaine users that also use several other substances.
Another strategy to modulate the amygdala-frontostriatal network is by using pharmacological agents. Currently there are no FDA approved pharmaceutical agents for the treatment of cocaine addiction, but medications that specifically act on the amygdala-frontostriatal circuitry might be promising novel treatment strategies. For example, corticotropin-releasing-factor (CRF) antagonists. The amygdala contains a high concentration of CRF terminals, cell bodies and receptors1 that are suggested to be involved in the negative reinforcement mechanisms that drive cocaine use20. Indeed, CRF antagonists, have shown to reverse stress-like responses during drug withdrawal and to reduce drug-self administration328,346. Other promising pharmaceutical agents are those that target the noradrenergic stress-system. For example, propranolol has shown to reverse cocaine withdrawal related cognitive-inflexibility347 and anxiety348, although not all studies using propranolol in the treatment of cocaine addiction are effective167.
It could very well be that interventions that target the amygdala-frontostriatal circuitry, such as rTMS, CRF-antagonists or behavioral therapy, may not be effective in all cocaine users. However they are likely to be effective in those that suffer most from negative emotional states and those with deficient emotion regulation. Therefore, increasing knowledge on the amygdala and its interaction with the frontostriatal system in substance use disorder will hopefully improve treatment outcome in those that are currently treatment-resistant.