In this study, we identified six functional networks related to the MCL system (VTA, NAcc, MD thalamus, amygdala, hippocampus and rACC) based on synchronized resting-state BOLD signal fluctuation. These functional networks involve multiple cortical and subcortical regions that are known to engage in reward processes, learning, memory and emotional regulation and are generally consistent with the known anatomical connections between these regions. When compared with matched healthy controls, cocaine users demonstrate similar network connection but with almost universally reduced connectivity strength for every MCL region studied except the NAcc. In contrast, while the primary motor, auditory and visual cortex seeds show connectivity with expected cortical and thalamic regions, there was no significant difference between the CU and HC group maps when the seeds were in primary motor or auditory cortex; the visual cortex seed did show increased connectivity to bilateral fusiform and lingual gyri in the CU group.
Although rsFC, by definition, does not directly speak either to brain activation during task performance or its correlation with task behavior, rsFC strength has been demonstrated to predict behavioral task performance requiring the use of that circuit (Hampson et al., 2006a
; Hampson et al., 2006b
; Kelly et al., 2008
; Seeley et al., 2007
). These observations suggest that synchronous fluctuations during the resting state are related to the ability of particular circuits to perform appropriately when challenged to perform a task. Indeed, Smith et al recently reported that resting networks identified using independent component analysis (a data driven analysis method) in 36 subjects demonstrated virtually identical networks as revealed in a meta-analysis consisting of nearly 30,000 subjects performing cognitive tasks categorized under one or more of 66 behavioral domain classifications, suggesting that functional networks utilized by the brain to perform virtually all cognitive processing are communicating even in the absence of formal task engagement (Smith et al., 2009
). Thus, the current results suggest circuit alterations throughout the MCL reward system that might underlie behavioral dysfunctions seen in cocaine dependent individuals.
In particular, the amygdala seed showed decreased rsFC in the CU group with mPFC, a connection known to be important for reversal learning (Chachich and Powell, 1998
). This construct has been shown to be deficient in cocaine addicts (Ersche et al., 2008
) and is thought to relate to user s difficulty in learning that cocaine cues may no longer signal reward after the development of problematic use. The hippocampus demonstrated decreased rsFC with mPFC, a region known to be involved in the processing of self-relevant information and emotional memories (Buckner et al., 2008
). Cocaine users demonstrate a notable lack of ability to recall negative consequences of past use and integrate such memories into decision-making processes (Washton, 1986
). Such difficulty in recall could also reflect difficulty in learning from negative outcomes. The reduced rsFC strength of the VTA – NAcc circuit could be reflective of an impaired Temporal Difference Error (TDE) dopamine signaling system, which has been shown to necessary for learning new stimulus-reward associations (Schultz, 2002
The MD thalamus seed yielded decreased cocaine group rsFC with extensive striatal regions. These circuits are thought to be important for focusing and maintaining desired behaviors while suppressing unwanted behaviors (Haber and McFarland, 2001
). Cocaine dependent individuals are known have difficulties with response inhibition (Kaufman et al., 2003
; Verdejo-Garcia et al., 2007
), which likely contributes to their propensity to relapse in the presence of cocaine-related cues. Finally, seeds in the rACC demonstrated decreased connectivity in CU subjects with the insula and amygdala, areas important for emotional functioning (Stein et al., 2007
). Emotional difficulties are characteristic of cocaine withdrawal (American Psychiatric Association., 1994
) and are highly comorbid with cocaine use disorders, although the current study excluded those who met diagnostic criteria for mood or anxiety disorders.
The lack of differences from the NAcc seed was unexpected. However, cocaine dependence has been hypothesized to relate to aberrant habit formation. If true, then altered connectivity, most likely increased, might be expected in dorsal, rather than ventral striatum/NAcc, as a shift in striatal circuits from ventral to dorsal has been demonstrated as drug dependence progresses (Porrino et al., 2007
Ma et al. conducted a study similar to ours in heroin dependent individuals, most of whom were maintained on methadone (Ma et al., 2010
). They reported increased functional connectivity from NAcc and amygdala to various frontal regions and decreased functional connectivity between frontal regions involved in executive function. An important difference between the two studies, aside from the obvious difference in drug use, is that their participants were mostly under the acute effects of a drug, i.e., methadone. As pointed out in their discussion, methadone can enhance responses to drug cues (Curran et al., 1999
; Langleben et al., 2008
) along with profound effects on mu opiate receptor circuits (Gray et al., 2006
). The difference between our results may, therefore, relate to the acute drug effects present in their study. Further examination of cocaine users under the influence of acute cocaine and heroin users not in methadone treatment will be needed to resolve these discrepant results.
While none of our participants were acutely intoxicated, twenty-one had positive urine tests for cocaine, indicating use within the last 3–4 days. Recent use was associated with decreased connectivity between the amygdala seed and left posterior insula and with increased connectivity between the primary motor cortex seed and left anterior insula. However, these recency affected regions did not overlap with those regions showing decreased cocaine group rsFC. While the insula is important for processing information on internal bodily sensations that are associated with cocaine use and has been implicated in craving, the specific significance of connectivity from amygdala and motor cortex to insula awaits further investigation.
Our results are in general agreement with findings of reduced gray matter density in medial frontal, rACC and insula in cocaine users (Franklin et al., 2002
) and with reduced white matter integrity in inferior frontal regions, also in cocaine dependent individuals (Lim et al., 2008
). Gray matter density reductions could result in reduced functional connectivity, depending on the cell populations responsible for the density difference. To the extent that functional connectivity reflect anatomical connections (Krienen and Buckner, 2009
), reduced white matter integrity would likely result in reduced functional connectivity.
From a theoretical standpoint, our results are most consistent with the hypothesized reduction in hedonic set point (Koob and Le Moal, 1997
). If rsFC strength is related to the ability to engage a circuit when required to perform a task, reduced connectivity throughout the MCL system would suggest a greater than normal stimulation requirement in order to engage reward (and other cognitive) pathways, thus leading to a shift from interest in and the ability of natural rewards to activate the system to the need for supra-physiological, drug-induced rewards to engage MCL circuits. This hypothesis could be directly tested in users in the absence and presence of acute cocaine administration.
Recent evidence strongly points to functional connectivity as reflective of and constrained by underlying neuroanatomical pathways (Krienen and Buckner, 2009
). In the present study, the VTA seed connectivity pattern demonstrated the expected extensive relationship with the striatum and prefrontal regions. Somewhat less expected was the strong connectivity seen between the VTA and the thalamus in both HC and CU groups, with large portions of thalamus significantly different between groups. While the VTA is known to project to numerous thalamic nuclei including the MD nucleus, the function of these projections has received little attention in the literature (Oades and Halliday, 1987
), despite evidence that MD thalamus lesions reduce cocaine self-administration in rats (Weissenborn et al., 1998
). The thalamus has recently been reported to be hypoactive in cocaine users performing a visual attention task, a result hypothesized to be related to attention and perception difficulties in addicts (Tomasi et al., 2007
). In light of these findings, it would seem that thalamic functions in cocaine dependent individuals warrant further investigation.
An important issue in interpreting results of a cross-sectional study such as ours is whether differences between groups are a consequence of chronic drug use or alternatively, reflect pre-existing differences that predispose some individuals to addiction. The preclinical literature provides extensive support for the notion that chronic cocaine use produces long-lasting alterations in MCL circuits and indeed, provided the inspiration for our choice of seed regions (Koob and Le Moal, 2001
). However, considerable data also point to genetic influences in cocaine dependence (Kreek et al., 2005
; Saxon et al., 2005
), which necessarily predate the effects of cocaine. As a first step in identifying circuits that have been changed by chronic cocaine use, we correlated connectivity strength with both years of cocaine use and current cocaine usage. A significant negative correlation with years of cocaine use was found between the VTA seed and bilateral thalamus/left lentiform nucleus/right NAcc. Critically, those individuals with fewer years of use showed connectivity strength in the range of healthy controls while those with more years of use showed weaker connectivity strength. Since reduced connectivity is thought to indicate more difficulty volitionally engaging a circuit when needed, this observation appears consistent with the clinical observation that cocaine dependent individuals seek out the supraphysiological stimulation of cocaine over natural rewards (American Psychiatric Association., 1994
). Preclinical studies also support the stronger reward activating capacity of cocaine; when given a choice, rats will self-administer cocaine over natural rewards (Grigson and Twining, 2002
). Further, only cocaine self administration, but not passive cocaine or food, can induce a persistent VTA synaptic enhancement that is resistant to behavioral extinction, which may represent a fundamental phenomenon driving pathological drug-seeking behavior (Chen et al., 2008
). Although the number of years of cocaine use has not been shown to be a predictor of treatment outcome (Poling et al., 2007
), these data may still speak to important neuroadaptations that characterize the clinical condition. This above conjecture requires further confirmation, perhaps using reward task activation and choice paradigms in cocaine users,
While most human imaging studies report altered limbic and frontal regions in cocaine users, a possible role for primary sensory and motor region involvement in cocaine dependence has been previously suggested. Kosten et al (Kosten et al., 2006
) report that relapse to cocaine use is associated with increased activation to cocaine cues in sensory association and motor cortical areas, along with posterior cingulate cortex. Tomasi et al (Tomasi et al., 2007
) showed that cocaine users had increased occipital activation in a sustained visuospatial attention task. While we saw no differences in connectivity strength between the HC and CU groups when seed regions were placed in motor or auditory cortex, we did find increased connectivity between a visual cortex seed with bilateral fusiform and lingual gyri. The significance of these observations is not clear, as extensive behavioral characterization of our subjects requiring visual processing was not performed. Nevertheless, this result, taken together with those cited above, support the importance of understanding plastic changes in brain function between cocaine users and healthy control individuals outside traditional reward related circuits. Finally, the null results with seeds in the primary motor and auditory cortex along with the enhanced visual seed connectivity in the CU group support the selectivity and specificity of the MCL-limited connectivity reductions observed herein.
Several limitations of this study warrant discussion. It should be emphasized that circuit directionality between nodes cannot be determined from rsFC analysis and is thus not implied in these data. The question of directionality of influence in affected circuits will need to be investigated with other techniques. It also should be noted that the presence of direct anatomical connections should not be concluded from functional connectivity analysis, although as noted, evidence strongly supports this assumption. However, as cocaine affects all three major monoamine systems (dopamine, norepinephrine and serotonin) which are thought to, at least in part, have a modulatory neuronal processing function, our observed differences in functional connectivity may relate to altered common input from any one or more of these systems.
Another limitation is that participants were not directly debriefed on what they were thinking about during the resting scan. While no one in the CU group showed overt signs of craving, it is possible that common thought patterns among the addicts could have contributed to the differences seen in this study. Finally, while there was no significant difference in number of dependent smokers in the CU and HC groups, there were more casual smokers among the CU group, reflecting the relatively common occurrence that many cocaine users smoke only when using cocaine. Therefore, it is difficult to completely disentangle effects related to nicotine from those related to cocaine. Insofar as drugs of abuse are thought to utilize a final common pathway involving MCL structures, it might be that our results reflect differences common to both nicotine and cocaine. The lack of differences from MCL seeds when comparing dependent smokers to the rest of the cohort supports the conclusion that our results are largely due to properties unique to the cocaine users. Further studies are needed to address fully this issue.
To the best of our knowledge, this study reports the first circuit level abnormalities in human cocaine users and demonstrates widespread reductions in the connectivity of multiple MCL system components, implying possible difficulty in appropriately activating reward, learning and emotional circuitry in cocaine dependent individuals. It would be of great interest to further investigate how plastic these apparent neural adaptations are and their potential candidacy as biomarkers that can be used to assess treatment matching and outcome prediction. Prospective studies are recommended to explore the relationship between connectivity within MCL pathways and specific behavioral features of addiction.