A shared spatial topography links the functional connectome correlates of cocaine use disorder and dopamine D_2/3 receptor densities

Abstract

Background:

The biological mechanisms that contribute to cocaine and other substance use disorders involve an array of cortical and subcortical systems. Prior work on the development and maintenance of substance use has largely focused on cortico-striatal circuits, with relatively less attention on alterations within and across large-scale functional brain networks, and associated aspects of the dopamine system. The brain-wide pattern of temporal co-activation between distinct brain regions, referred to as the functional connectome, underpins individual differences in behavior. Critically, the functional connectome correlates of substance use and their specificity to dopamine receptor densities relative to other metabotropic receptors classes remains to be established.

Methods:

We comprehensively characterized brain-wide differences in functional connectivity across multiple scales, including individual connections, regions, and networks in participants with cocaine use disorder (CUD; n=69) and healthy matched controls (n=62), Further, we studied the relationship between the observed functional connectivity signatures of CUD and the spatial distribution of a broad range of normative neurotransmitter receptor and transporter bindings as assessed through 18 different normative positron emission tomography (PET) maps.

Results:

Our analyses identified a widespread profile of functional connectivity differences between individuals with CUD and matched healthy comparison participants (8.8% of total edges; 8,185 edges; p_FWE=0.025). We largely find lower connectivity preferentially linking default network and subcortical regions, and higher within-network connectivity in the default network in participants with CUD. Furthermore, we find consistent and replicable associations between signatures of CUD and normative spatial density of dopamine D_2/3 receptors.

Conclusions:

Our analyses revealed a widespread profile of altered connectivity in individuals with CUD that extends across the functional connectome and implicates multiple circuits. This profile is robustly coupled with normative dopamine D_2/3 receptors densities. Underscoring the translational potential of connectomic approaches for the study of in vivo brain functions, CUD-linked aspects of brain function were spatially coupled to disorder relevant neurotransmitter systems.

Introduction

The study and treatment of substance use disorders represents a complex, multifaceted challenge with far-reaching implications for individuals, their families, and our broader society. In particular, increasing prevalence of cocaine use disorder (CUD) substantially contributes to the rising overdose deaths in the United States (1). A fundamental question facing the field of addiction neuroscience concerns the extent to which substance use behaviors emerge through local patterns of activity or are instantiated across the broader large-scale networks of the human brain. While prior foundational work has established cortico-striatal-thalamic circuit disruption as a fundamental feature of substance use disorders (2), consistent with systems-level models of substance use disorders (3), striatal circuitry is deeply embedded within spatially distributed and functionally linked systems that span the cortical sheet. Whether alterations in functioning are isolated to specific circuits or diffusely distributed throughout large-scale network architecture remains largely unexplored.

Cocaine preferentially targets the dopamine system, and both tonic and phasic dopamine neurotransmission have been shown to play a critical role in the onset and maintenance of substance use pathology (4). Here, for instance, reduced activity within the large-scale networks supporting attention and inhibitory control points to an imbalance between the core dopaminergic circuits that underlie subjective valuation and conditioned responding and those that support “higher-level” executive functioning. Moreover, the neuromodulatory impact of cocaine is not specific to the dopamine system, while primarily blocking the dopamine transporter and inhibiting its reuptake from the synaptic cleft, it also modulates serotonin and norepinephrine transporters (5). However, the extent to which the brain functional correlates of CUD may be coupled to the spatial distribution of dopaminergic processes, relative to other neurotransmitters, remains to be established.

Here, we investigate the relationship between CUD, whole-brain functional connectivity, and neurotransmitter receptor densities. First, we used the network-based statistic (6) to derive whole-brain functional connectivity differences between individuals with CUD and controls. We then examine the association between the identified functional network and the spatial distribution of receptor densities, inferred from positron emission tomography (PET). In doing so, we demonstrate preferential correspondence between regional connectivity alterations related to CUD and the normative topography of dopamine D_2/3 receptor densities across three independent PET datasets. These data reliably establish that in CUD, extensive and brain-wide alterations in connectivity exist and are closely coupled with dopamine D_2/3 receptor densities.

Methods

Participants

The current study used data from the SUDMEX CUD imaging dataset (7). A total of n=131 individuals (age range: 18–50), including 69 individuals with CUD (85.51% male) and 62 demographically matched healthy comparison participants (79.03% male) were included in the present study. Notably, these data represent a diverse and non-European-centric population in Mexico City, Mexico. Participants with CUD had to have used for at least one year, with current average use of at least three times per week, with periods of continuous abstinence of less than one month during the last year. Additional participant inclusion criteria can be found in Supplementary Section 1. Participant behavioral characteristics and demographics can be found in Table 1. The reported study analyses procedures were approved by the Yale University Institutional Review Board IRB #1507016245.

Table 1:

Demographic characterization of study sample (n=131).

Group	Cocaine Use Disorder (CUD)	Healthy Comparison
Participants (n, % Male)	69(85.51)	62(79.03)	χ2: 0.30 p=0.583
Age (mean±sd)	31.34±7.27	30.42±8.18	t: −0.68; p=0.501
Education (mean±sd)	2.83±1.27	3.52±1.39	χ2: 13.40 p<0.020*
Head Motion(mean±sd)	0.23±0.10	0.21±0.08	t: 1.63; p=0.105

Three participants were excluded for missing age values. Seven participants were excluded for missing education values. Three participants (2 two HC, 1 CUD) missing sex values, male/female were the only available options. χ2, chi-square. Head motion calculated using mean framewise displacement (mm).

^*= significant group difference (p< 0.05).

MRI acquisition and processing

Intrinsic (resting state; fcMRI) functional imaging data were acquired using a 3T Phillips Ingenia MR scanner in Mexico City, Mexico. Field-standard processing and quality control procedures were implemented. To generate whole-brain functional connectivity matrices, we parceled each individual’s normalized scans into 400 cortical (8) and 32 subcortical (9) regions. (Fig. 1A). Further details can be found in in Supplement Section 3.

Figure 1.

Whole brain atypical functional connectivity in cocaine use disorder (CUD). A widespread network of affected connections exists between individuals with CUD and healthy matched controls, extending across the functional connectome. A) Schaefer 7-network and Tian subcortex parcellations (Scale II) from left to right: a indicates anterior; AMY, amygdala; CAU, caudate nucleus; d, dorsal; DA, dorsoanterior; Default, default network; DorsAttn, dorsal attention network; DP, dorsoposterior; FPN, frontoparietal network; GP, globus pallidus; HIP, hippocampus; l, lateral; Lim, cortical limbic network; m, medial; MTL, medial-temporal lobe (amygdala and hippocampus); NAc, nucleus accumbens; p, posterior; SomMot, somatomotor network; Stri, striatum; PUT, putamen; THA, thalamus. B) Images with a red color scale represent number of significant edges (degree) where individuals with CUD show hyperconnectivity. B) Images with a red color scale represent number of significant negative edges of NBS network where individuals with CUD show hypoconnectivity. C) Heatmap quantified using raw total edge count (upper triangle) and normalized proportion of edges based upon network size (lower triangle) within the NBS component that fall within each of the canonical networks. The darker red indicates higher connectivity in CUD. D) Images with a blue color scale represent number of significant negative edges of NBS network where individuals with CUD show hypoconnectivity. E) Heatmap quantified using raw total edge count (upper triangle) and normalized proportion of edges based upon network size (lower triangle) within the NBS component that fall within each of the canonical networks. Darker blue color indicates lower connectivity in CUD

Whole-brain functional connectome correlates of cocaine use disorder

Non-parametric ANCOVA models were used to analyze brain-wide functional connectivity differences between individuals with CUD and matched controls, adjusting for age, sex, and education. The Network Based Statistic (NBS) was used to perform familywise error-corrected (FWE) inference at the level of connected components of edges (12,13), with the primary component-forming threshold, τ, set to τ < .05 and significance assessed at p_FWE < 0.05. Further statistical details and results for τ = 0.01 and τ = 0.001 are reported in Supplementary Table 2 and Supplementary Section 4.

Associations between functional dysconnectivity and receptor densities

In order to investigate the relationship between functional alterations identified in individuals with CUD and the topographic distributions of normative neurotransmitter expression in healthy participants, we used Spearman correlation to examined spatial associations between the number of significant connections and normative receptor bindings across each of the 432brain regions. These associations were first assessed using 17 unique spatial maps that index a specific receptor or transporter with the largest available sample size (10, 11). Multiple maps from independent datasets were available for some of the receptors and transporters, using either the same or unique tracer. If available, these additional maps were used to assess the stability and replicability of any statistically significant associations (p < 0.05). Permutation-based inference (10,000 permutations) using ‘spin-tests’ were used to assess significance, while accounting for spatial autocorrelation. Further statistical details and information regarding specific tracers are provided in Supplementary Section 1 and Table 2.

Results

Wide-spread connectivity alterations in cocaine use disorder

We find a significant wide-spread pattern of both hyperconnectivity and hypoconnectivity associated with CUD, encompassing 8.8% of the total edges (8,185 edges; p_FWE=0.025) linking 432 brain regions (Fig. 1). The majority of significant edges (58.94%; 4,824 total edges) demonstrated hypoconnectivity in individuals with CUD. Here, the highest proportion of hypoconnected edges preferentially implicated the default network (Fig. 1D–E). After accounting for network size (see Supplemental Section 4), connections within striatum and thalamic regions, and between striatum and control networks were preferentially implicated in participants with CUD (Fig. 1D–E). At a regional level, precuneus posterior cingulate cortex, medial prefrontal cortex, and anterior caudate nucleus were among the areas most strongly implicated in the network of lower functional connectivity.

When considering patterns of higher connectivity in the CUD group, hyperconnected edges accounted for 41.06% of the total significant edges (3,361 total edges). The total number of hyperconnected edges demonstrated preferential within-network connectivity of the default network, as well as between-network connectivity of striatum and ventral attention networks (Fig. 1B–C). When normalizing for the total size of a given network, between-network hyperconnectivity of the striatum-somatomotor networks preferentially emerged. At a regional level, frontal operculum, parietal operculum, extrastriate cortex, and anterior putamen, were among the areas most strongly implicated in the network of higher functional connectivity.

Shared spatial topography links cocaine use disorder and D_2/3 receptor densities.

Regional functional dysconnectivity was significantly correlated with D_2/3 receptor density ([¹¹C]FLB 457, ρ=0.175 p_spin=0.015). Associations with D_2/3 receptors replicated across two additional normative PET maps ([¹⁸F]fallypride, ρ=0.168; p_spin=0.022) and ([¹¹C]FLB 457, ρ=0.192; p_spin=0.007) (Fig. 2B–C), indicating robust and reliable relationships between D_2/3 receptors density and CUD-related connectivity dysfunction (Fig. 2). To ensure that this association was not driven by large differences in tracer binding between cortical and subcortical regions, we replicated the D_2/3 receptors association after excluding subcortical regions (Supplementary Fig. 1). Associations with two serotonin results were also significant (5HT₄ [¹¹C]SB207145: ρ=0.143; p_spin=0.032, and 5HT₆ [¹¹C]GSK215083: ρ=0.136; p_spin=0.020) and reported in Supplementary Fig. 3, but did not have replication samples. No associations with other available neurotransmitter systems were detected (Supplementary Table 2).

Figure 2.

Spatial overlap between whole-brain Network Based Statistic (NBS) network and D_2/3 receptor density in cocaine use disorder (CUD). A) Visualization of the total (positive and negative) number of significant edges at each region within the NBS component (Fig. 1B + 1D) where change in fcMRI was significantly correlated with the spatial D_2/3 receptor density in a discovery sample (Sandiego 2015, p_spin=0.019) and two replication samples (Jaworska 2020, p_spin=0.030 and Smith 2017, p_spin=0.013), respectively). B) D_2/3 binding potential of PET samples for each receptor source, i.e., discovery sample and replication samples. Color scale normalized between −1.0 to 1.0 for cortex and subcortex separately. C) Each violin-box plot contains (from left to right) distribution of 10k spin-test null correlations between each edge of the NBS component and the spatial density of D_2/3 receptors. Red dot indicates significant spearman’s correlation. * reflects statistical significance at the threshold p_spin<0.05. Discovery Sample: Sandiego et al., (2015) (12); Replication 1: Jaworska et al., 2020 (13); Replication 2: Smith et al., 2017 (14).

Discussion

Cocaine use disorder (CUD) emerges, in part, through the complex interactions of biological systems encompassing neurochemical cascades and associated functional interactions across both local circuits and broader large-scale networks. Establishing how these processes contribute to the onset and maintenance of substance use disorders requires a multi-scale approach, considering measures of in vivo brain function, as assessed through fcMRI, as well as neurotransmitter synthesis and transport assessed though PET imaging. In the present analyses, we find wide-spread alterations in intrinsic (“resting-state”) functional connectivity in CUD and by integrating these findings with PET data, we demonstrate the presence of shared spatial patterns linking D_2/3 receptor densities with the functional connectome correlates of CUD.

While prior work has revealed disruptions in cortico-striatal-thalamic circuitry that underlie varying stages in of substance use disorders (2), our findings support a more diffuse, brain-wide dysregulation in CUD, extending the beyond neural circuit-specific hypotheses. In addition to striatal and thalamic regions, we find alterations in large-scale cortical networks, including the default mode, control, somatomotor, and ventral attention networks, suggesting that dysfunction extends beyond atomically constrained cortico-striatal-thalamic circuitry.

Critically, our findings demonstrate a reliable spatial correspondence between functional dysconnectivity in CUD and the dopaminergic system, extending across both cortical and subcortical regions. Cocaine acts by binding to the dopamine transporter, blocking the reuptake of dopamine from the synaptic cleft, as well as blocking the transporters for norepinephrine and serotonin (5). While our findings also implicate parts of the serotonin system, the most replicable and robust link was found with D_2/3 receptor densities, suggesting that brain dysconnectivity across large-scale brain networks are preferentially coupled to dopaminergic pathways.

Prior investigations have demonstrated distinguishable functional connectome profiles between CUD and other substance use disorders, such as opioid use disorder, suggesting that individual variability in large-scale connectomes may serve as a valuable predictor for treatment outcomes in CUD (15). While our findings demonstrate a robust pattern of functional alteration in CUD that is coupled with dopaminergic pathways, the extent to which the present findings may reflect a substance specific neurobiological profile or a general profile related to dopaminergic drugs of misuse (for instance, opiates, alcohol, and cocaine) remains to be determined.

The current study, similar to many neuroimaging datasets of individuals with substance use disorders, is limited by its cross-sectional nature and longitudinal approaches may provide further insight on whether neurobiological profiles reflect a vulnerability for illness, a direct consequence of substance use, or the biproduct of illness linked environmental impacts. Moreover, further investigations using concurrent PET and fMRI imaging in patient samples is needed to determine whether illness-related neurochemical alterations interact with brain function. The present sample is also characterized by a large proportion of male participants. Prior work has established the importance of sex differences in the brain-behavior features that characterize substance use disorders (16). Accordingly, data from more sex diverse individuals should be obtained in the future.

Conclusions

Together, these data establish correspondence across the functional networks implicated in CUD and the neurotransmitters that underlie its mechanism of action. This provides a foundation for future work disentangling the biological mechanisms that govern individual variances in the dopaminergic systems, functional brain organization, and substance use.

Key Points.

Question:

Are there group differences in whole brain functional connectivity between individuals with and without cocaine use disorder, and to what extent do these connectivity patterns relate to the spatial distribution of dopamine (D_2/3) receptor densities?

Findings:

The presence of cocaine use disorder is associated with brain-wide functional connectivity alterations that are spatially coupled to the density of dopamine (D_2/3) receptors.

Meaning:

A preferential and replicable link exists between the functional connectome correlates of cocaine use disorder and dopamine receptor densities across the brain.