Transcranial Direct Current Stimulation Applied to the Dorsolateral and Ventromedial Prefrontal Cortices in Smokers Modifies Cognitive Circuits Implicated in the Nicotine Withdrawal Syndrome

Abstract

Background:

The Nicotine Withdrawal Syndrome remains a major impediment to smoking cessation. Cognitive and affective disturbances are associated with altered connectivity within and between the Executive Control Network (ECN), Default Mode Network (DMN), and Salience Network. We hypothesized that functional activity in cognitive control networks, and downstream amygdala circuits, would be modified by application of transcranial direct current stimulation (tDCS) to the left (L) dorsolateral prefrontal cortex (dlPFC, ECN) and right (R) ventromedial prefrontal cortex (vmPFC, DMN).

Methods:

15 smokers (7 women) and 28 matched nonsmokers (14 women) participated in a randomized, sham-controlled, double-blind, exploratory crossover study of three tDCS conditions: Anodal-(L)dlPFC/Cathodal-(R)vmPFC, reversed polarity, and sham. Cognitive tasks probed withdrawal-related constructs (error monitoring; working memory; amygdala reactivity), while simultaneous fMRI measured brain activity. We assessed tDCS impact on trait (nonsmokers vs. sated-smokers) and state (sated vs abstinent) smoking aspects.

Results:

Single-session, Anodal-(L)dlPFC/Cathodal-(R)vmPFC tDCS enhanced deactivation of DMN nodes during the working memory task and strengthened anterior cingulate cortex activity during the error monitoring task. Smokers were more responsive to tDCS-induced DMN deactivation when sated (vs. withdrawn), and displayed greater cingulate activity during error monitoring than nonsmokers. Nicotine withdrawal reduced task engagement and attention, and reduced suppression of DMN nodes.

Conclusions:

Cognitive circuit dysregulation associated with nicotine withdrawal may be modifiable by anodal tDCS applied to L-dlPFC and cathodal tDCS applied to R-vmPFC. tDCS may have stronger effects as a complement to existing therapies, such as nicotine replacement, due to possible enhanced plasticity in the sated state. (NCT01511614, https://clinicaltrials.gov/).

Introduction

The Nicotine Withdrawal Syndrome (NWS) significantly contributes to the high rate of smoking relapse (1,2). NWS peaks within one week of quitting, is associated with craving, negative affect (dysphoria, irritability, anxiety), attentional and cognitive impairment (poor concentration, impulsive decision making), and is alleviated by acute smoking (3,4). Despite the well-known importance (5) and challenge (6,7) of smoking cessation, response rates to current FDA-approved pharmacotherapies remains poor; the most efficacious, varenicline, has ~27% absolute cessation rate (6-months post-quit), while various nicotine replacement therapies (e.g. patch, gum, nasal spray) fare worse (8,9). Inadequate therapeutic response may be explained by the neuroplastic nature of the disease; pharmacotherapy relieves withdrawal state symptoms, but acts minimally on underlying disease trait pathology. Addiction, a chronic, relapsing cycle of binge, intoxication, and withdrawal (10), develops from drug-induced neuroplastic and allostatic changes to reward, cognitive and self-referential brain circuits (11–14). Thus, a treatment acting directly on these circuits, such as noninvasive neuromodulation, would be of important clinical value. As such, transcranial direct current stimulation (tDCS) has been considered as an adjuvant or alternative treatment for smoking cessation (15–17). Here, we investigated the influence of tDCS on large-scale brain networks associated with smoking dependence and withdrawal.

The brain is intrinsically organized into multiple large-scale functional networks that appear to be stable across healthy individuals (18–22). Because networks are functionally (vs. anatomically) defined using data-driven methods, the total number of networks and their components can vary across parcellation methods. However, three networks have consistently emerged that are considered to be especially vulnerable to disruption in neuropsychiatric disorders (23–26): (1) the Executive Control Network (ECN), a “task-positive” network with core nodes in the dorsolateral prefrontal cortex (dlPFC) and posterior parietal cortex (PPC), which supports goal-directed performance and attentional control (27); (2) the Default Mode Network (DMN), localized within the ventromedial prefrontal cortex (vmPFC), parahippocampal gyrus (PHG) and the posterior cingulate cortex (PCC), which supports many forms of self-referential thought, including planning and rumination (28–30), and is generally anticorrelated to the ECN (20,31); and (3) the Salience Network (SN), thought to guide attention to endogenous or exogenous stimuli based on homeostatic and environmental context, and includes the anterior insula and dorsal anterior cingulate cortex (dACC) (32). Aspects of each of these networks downregulate the amygdala and related limbic circuitry (33).

The cognitive and affective disturbances of the NWS have been attributed to reduced circuit strength of the ECN and SN, increased connectivity strength of the DMN, and hyperactivity of the amygdala (34,35). This model has been supported by functional magnetic resonance imaging (fMRI) studies demonstrating network dysfunction as a trait characteristic of smoking, and a state characteristic of acute nicotine deprivation (36–42). Nicotine deprivation in smokers enhances amygdala reactivity (41), reduces anti-connectivity between the DMN and ECN (43), and enhances insula–DMN connectivity (44).

tDCS has the potential to modify neuronal circuits by applying a subthreshold electrical current through the scalp. tDCS alters resting functional networks (45–50), brain neurochemistry (51), and BOLD signal (52–55), and has been reported to reduce cigarette craving and consumption (56–61). We assessed acute, single-session tDCS modulation of the three large-scale brain networks and their associated cognitive functions in a cohort of healthy controls and nontreatment seeking smokers, in both nicotine sated and deprived states.

We hypothesized that function of the ECN, DMN, and SN would be modified by acute, single-session application of tDCS to cortical nodes of ECN (dlPFC) and DMN (vmPFC) (Fig. S1). We predicted that, compared to sham, excitatory, anodal-dlPFC tDCS paired with inhibitory, cathodal-vmPFC tDCS would enhance ECN and SN nodes, and downregulate DMN and amygdala nodes. Further, we predicted that the direction of network change would be inverted following tDCS polarity reversal.

Methods and Materials

Participants

Healthy, right-handed subjects, 18–60 years, were enrolled from the Baltimore, Maryland area (Fig. S2 for CONSORT diagram, Table 1 for demographics): an exploratory sample of 15 smokers (smoking ≥1 year with NicAlert ≥4) and 28 demographically matched nonsmokers (no past-year nicotine use and no lifetime daily nicotine use of >1 month). Exclusion criteria included MRI contraindications, major medical, neurologic or psychiatric conditions, regular medication use that might interfere with BOLD signal, and drug dependence (except nicotine in smokers). Written informed consent was obtained in accordance with the National Institute on Drug Abuse - Intramural Research Program (NIDA-IRP) Institutional Review Board.

Table 1:

Participant Demographics.

Total Completers (N = 43)	Nonsmokers (N = 28)	Smokers (N = 15)	p-value*
Gender (F:M)	14:14	7:8	1.00
Age (Mean: SD)	39.3 (10.3)	40.1 (12.0)	0.82
Years of Education (Mean: SD)	14.1 (2.1)	13.3 (1.63)	0.19
Highest Degree			0.46
Some H.S.	1	1
High School	6	4
Some College	9	8
A.A. Degree	1	1
College Graduate	9	1
Masters	2	0
WASI Full-4 IQ	101 (9.75)	103 (15.5)	0.59
Race			0.40
Asian	1	0
White	10	6
Black/African-American	15	6
Multiracial	1	3
American Indian/Alaska Native	1	0
Ethnicity (Hispanic: Not Hispanic)	2:26	1:14	1.00

^*Welch’s Two-Sample t-test for numerical data (R, stats::t.test), Pearson’s Chi-squared test for categorical data (R, stats::chisq.test, with simulated p-values for Education, Race, and Ethnicity).

Experimental Design

The randomized, double-blind, sham-controlled crossover design (Fig. 1A) involved three tDCS-fMRI sessions (detailed below). Smokers completed all three tDCS-fMRI sessions twice, across 2 days separated by ≥48h, randomized and blinded between a 24h-release nicotine (NicoDerm-CQ; GlaxoSmithKline) or placebo patch, each following biochemically verified 12h nicotine abstinence. Nicotine patch dose matched cigarettes smoked per day (‘cpd’; 10–15cpd, 21mg; 15–20cpd, 28mg; 20–25cpd, 35mg; ≥25cpd, 42mg), and was applied the morning of the appointment, ≥2h prior to scanning (62). Primary tDCS outcomes were behavioral and neural responses to three cognitive tasks, during tDCS-fMRI sessions.

Figure 1: Experimental Design, tDCS Montage and Simulation of tDCS-induced E-field Distribution.

(A) Experimental Design. Double-blinded, randomized, sham-controlled crossover study completed by both smokers (“Smoker timeline”) and nonsmokers (“Nonsmoker timeline”). All subjects completed three tDCS-fMRI sessions, randomized between the three tDCS conditions (part B of this figure). Smokers completed each visit following biochemically verified (CO monitor) 12-hours overnight nicotine abstinence, and placed a study patch, containing either nicotine or placebo, in the morning at least 2 hours prior to the first scan of the day. All conditions (study patch, tDCS condition) were randomized and counterbalanced. On the second visit day, the tDCS conditions were re-randomized for smoker participants. Subjects completed multiple simultaneous fMRI scans before, during and after tDCS, in a fixed order (“tDCS-fMRI session: scan order”). tDCS sessions were separated by at least a 90-minute inter-tDCS period (equating to an inter-scan interval – the time spent outside of the scanner between the end of one MRI session and start of another – of at least ~30–45 minutes). Each tDCS-fMRI scan lasted between 60 and 90 minutes. The study visit day (with 3 tDCS-fMRI sessions) lasted ~7–8 hours, including pre-MRI nursing assessments, task training and questionnaires. In some cases of equipment or participant scheduling issues, some tDCS sessions were separated by a number of days. The counterbalance conditions were determined by the NIDA-IRP Pharmacy and kept blinded there, with tDCS setting programmed by the MR-operator according to a key. Note that subjects performed the N-back twice, both during and after stimulation. Here, we report only the N-back online (during). A DTI scan was taken once per participant, at the end of one of the tDCS-fMRI scan sessions. Note: “tDCS-fMRI session: scan order” panel applies to all tDCS-fMRI sessions.

(B) tDCS Montage (2mA, 25 minutes) with three conditions, from left to right: (1) Anodal-(L)dlPFC + Cathodal-(R)vmPFC, “An-dlPFC”; (2) Cathodal-dlPFC + Anodal-vmPFC, “An-vmPFC”; and (3) Sham. tDCS was administered inside the MRI scanner with simultaneous fMRI scanning. The electrodes were 5cm × 7cm (35cm²) and covered evenly with a 3–5mm layer of conductive paste (Ten20, Weaver and Company). The (R)-vmPFC electrode was placed over the R-SOR parallel to the brow, and the (L)-dlPFC electrode was placed over F3, perpendicular to the brow.

(C) Across all 43 subjects, average normal component of the electric field, |E|_normal, for, from left to right: (1) anodal-(L)dlPFC/cathodal-(R)vmPFC stimulation, and (2) cathodal-(L)dlPFC/anodal-(R)vmPFC stimulation. Positive values denote inward current flow and negative values denote outward current flow. (D) Average of the E-field strengths across the 43 subjects on the cortical surface. The E-field magnitude distributions are the same for anodal and cathodal tDCS; however, the directions of current flow are opposite. L = Left, R = Right, dlPFC = dorsolateral prefrontal cortex, vmPFC = ventromedial prefrontal cortex.

Cognitive Tasks

We used three tasks to test circuitry implicated in the NWS: an N-back (Working Memory, “WM”) task, to measure ECN activation (63,37) and DMN suppression (20); a modified Parametric Flanker Task (error monitoring task) (36) to measure ACC and insular activity (SN nodes); and a modified Matching Faces and Shapes task (41) to measure amygdala reactivity, and, indirectly, prefrontal top-down control (64–66). Tasks were performed inside the scanner either during (online: N-back, 13m into tDCS) or immediately following stimulation (offline: Faces, 1m post-tDCS; Flanker, 10m post-tDCS). Task details are fully described in the Supplemental Methods. Each task has previously detected characteristics of nicotine dependence or withdrawal: N-back performance is predictive of smoking relapse (67); smokers demonstrate difficulty recruiting SN activity during the Flanker (36); and the amygdala is hyperactivated in NWS, whereas connectivity of amygdala, insula and DMN circuits is downregulated by nicotine agonism (40,41).

Transcranial Direct Current Stimulation

Subjects completed 3 tDCS sessions (2mA for 25m, 15s ramp; NeuroConn DC-Stimulator Plus-MR; neuroCare, München, Germany) per day within the MRI scanner -- two active, one sham -- separated by ≥90m washout (68). The order of tDCS sessions was randomized, counterbalanced, and double-blinded. In “An-dlPFC” tDCS, anodal stimulation was applied to L-dlPFC (Beam-F3 method (69)), and cathodal stimulation to R-vmPFC (R-supra-orbital ridge, Fig. 1B); polarity was reversed in “An-vmPFC” tDCS. For sham, current was briefly ramped on and off at session start, after which brief, minimal pulses were delivered only to provide an impedance reading (70,71). Participants completed a blinding questionnaire following each tDCS session. See the Supplemental Methods for additional tDCS details.

Image Acquisition

Whole-brain echo-planar functional images, structural T1-weighted brain images (MPRAGE) and diffusion tensor (DTI) images were acquired on a 3T Siemens Magnetom Prisma scanner at NIDA-IRP (Baltimore, Maryland) with a 20-channel head coil (Supplemental Methods, Table S1).

Statistical Analysis

We generated statistical models to test tDCS effects on two aspects of smoking dependence: Trait-Smoking, the between-subjects factor of nonsmokers vs. sated-smokers; and Nicotine-State, the within-smokers factor of nicotine abstinence vs. satiety. For Faces task only, due to technical issues, data for 1 smoker were lost in Trait-Smoking analyses (N_Smoker=14); an additional smoker did not respond to this task under the placebo patch, and their data were removed in Nicotine-State analyses (N_Smoker=13).

Behavioral Measures

We measured response speed (inverse Reaction Time, 1/s) and accuracy on each trial. On the N-back, sufficient hits, false alarms, and errors (commission/omission) occurred to evaluate signal detection theoretic measures (d-prime sensitivity, criterion, and omissions). Modeling was conducted using R Statistical Computing software (with packages tidyverse (72), afex (73), emmeans (74), neuropsychology (75)). Type III Repeated-measures ANOVA (afex::aov_ez) modeled three mixed factors for each task: within-subjects factor of tDCS stimulation (An-dlPFC, An-vmPFC, Sham); within-subjects factor of Task Level (i.e. difficulty on N-back and Flanker, Matching type on Faces); and between-subjects factor of Group for the Trait-Smoking model (Nonsmoker, Smoker-sated), or within-subjects factor of Patch for the Nicotine-State model (Nicotine, Placebo). We evaluated significant interactions (p<0.05) with Tukey post-hoc contrasts (emmeans).

Imaging Analysis

fMRI scans were pre-processed with the Brain Imaging Data Structure (76) application fmriprep (v1.3.1) (77)), while EPI brain-masking, first-level and group processing were carried out in AFNI (v.18.1.24) (78). Statistical tests on region of interest (ROI) data were carried out in R. We simulated the average electrical field (E-field) induced by tDCS across all subjects using MPRAGE and DTI scans, processed with SimNIBS v.3.0 (79). See Supplemental Methods for details.

Region of Interest Analysis

For each task, we pursued a ROI analysis to directly test the associated network-based hypotheses. For N-back, nine ECN-associated regions were derived from the Neurosynth (http://neurosynth.org) “working memory” map: bilateral Inferior Frontal Gyrus, bilateral Inferior Parietal Lobule/Angular Gyrus, bilateral Middle Frontal Gyrus/BA10 (MFG), R-MFG/BA6, and bilateral Cerebellum/Crus 1. For Flanker, three SN nodes were functionally defined (36): right ACC, and bilateral Insula. For Faces, six bilateral amygdala sub-regions were defined (80): superficial (corticoid) nuclei, centromedial groups, and laterobasal complexes. First-level (participant-session) models of correct-press trials were generated (AFNI:3dDeconvolve, 3dREMLfit), including regressors for six head motion parameters, incorrect and omitted events, and censoring timepoints with head motion >0.5mm Framewise Displacement (81). For N-back and Faces, first-level contrasts represented high WM-load (3-back minus 0-back) and amygdala reactivity (Faces minus Shapes). To maintain the Flanker’s parametric design, we generated first-level processed files for each difficulty level (i.e. number of flankers), which were then used as levels in a within-subjects, task-contrast factor for the second-level ANOVA (below). Next, we extracted BOLD signal from the ROIs for each task (AFNI:3dROIstats) across all sessions and subjects. We removed outlier data points (>1.5*Interquartile Ratio beyond 25^th or 75^th percentiles) for all values measured within the ROI.

We modeled factors of interest using a mixed-model ANOVA (afex::mixed). For N-back and Faces, we modeled two factors: (1) the within-subjects factor of tDCS; and (2) a between-subjects factor Group (Trait-Smoking model), or a within-subjects factor of Patch (Nicotine-State model). For the Flanker, we included a within-subjects factor of task difficulty to generate a parametric task contrast. We applied a Bonferroni-corrected significance threshold of p<0.05 for number of ROIs for each task (N-back, nine ROIs, p_raw<0.0057; Flanker, three ROIs, p_raw<0.0167; Faces, six ROIs, p_raw<0.0085).

Whole Brain Analysis

We performed exploratory whole-brain analyses using multifactorial ANOVA (AFNI:3dMVM). Main effects and interactions of tDCS, Group or Patch, were corrected for whole-brain familywise error (FWE) α<0.01 and p-voxelwise<0.001, with cluster size determined using AFNI:3dClustSim for each task/model dataset. Task maps were generated by assessing the main effect of task difficulty for the Flanker, and a first-level contrast for N-back (3-back minus 0-back) and Faces (Faces minus Shapes). We applied a stringent whole-brain correction of at least α<0.01 FWE for all task maps.

Lastly, we explored whether tDCS altered brain-behavior relationships using whole-brain regression (3dRegAna, Supplemental Methods).

Results

Results are organized by statistical model: Trait-Smoking, the between-subjects comparison of sated-smokers to nonsmokers; and Nicotine-State, the within-smokers comparison of sated vs. withdrawal state.

Task Behavior

Overall, the tasks produced expected behavioral outcomes (Fig. S3, S4).

Trait-Smoking Model:

Sated smokers performed more accurately, missed fewer responses, and had lower signal criterion than nonsmokers on the N-back (p<0.05, Fig.S3A, S5A), and performed faster, with an accuracy Group*Task interaction on Faces (p<0.05, Fig.S3C–D). There were no main effects or interactions of tDCS.

Nicotine-State Model:

Across all three tasks, smokers wearing the placebo (vs. nicotine) patch demonstrated slowed response speed (all p<0.05), and reduced accuracy on the N-back (p=0.003) and Faces (p=0.02) tasks (Fig. S4). Moreover, smokers on the placebo patch condition also demonstrated reduced d-prime sensitivity (p=0.01), increased criterion (p=0.01) and omissions (p=0.003) when performing the N-back task (Fig. S5B). There were no main effects or interactions of tDCS.

Region of Interest fMRI Analysis

Trait-Smoking Model:

There was a strong trend-level main effect of tDCS (p_corrected=0.05) that was driven by a tDCS*Group interaction (p_corrected=0.03) in the right ACC (Fig. 2), when combined across all task difficulty levels. An-dlPFC increased ACC activity above both An-vmPFC and Sham (post-hoc p_{An-dlPFC-Sham}=0.04, p_{An-dlPFC-An-vmPFC}=0.04) across all subjects. This effect was most pronounced in smokers (post-hoc p_{Smoker[An-dlPFC-Sham]}=0.005). There were no significant effects of tDCS or Group within the N-back or Faces ROIs.

Figure 2: Parametric Flanker Task, Trait-Smoking Model (ROI results).

(A) ROI mask of bilateral insula and right dACC, derived from Ref.(36). Arrow points to right dACC. (B) Main effect trend of tDCS in the right dACC (F _tDCS (2, 324.07) = 4.08, p_corrected = 0.05). (C) tDCS * Group (sated-smokers vs. nonsmokers) interaction on right dACC activation (F _tDCS:Group (2, 324.07) = 4.84, p_corrected = 0.03). Green = dACC, Red = Right Insula, Blue = Left Insula, R = Right, ACC = Anterior Cingulate Cortex, * < 0.05, + < 0.1, Error bars = SE.

Nicotine-State Model:

There were no tDCS or Patch effects on any task ROI.

Whole-brain Exploratory Analysis

Trait-Smoking Model:

All three tasks produced robust, expected activation maps (Fig. S6–S8, Tables S2A, S3–S4), with N-back further producing the expected high WM-load effect for the 3–0 contrast (vs. 1–0, Fig. S6B). There was a main effect of tDCS within nine regions associated with the DMN during high WM-load across all 43 subjects (e.g. hippocampus, PHG, superior temporal gyrus, inferior parietal lobule, mid-cingulate gyrus, and precuneus; FWE corrected α<0.01; Fig. 3A, Table S2B). There were no effects of Group. Additionally, An-dlPFC tDCS suppressed BOLD signal in the DMN-related regions more than Sham and An-vmPFC tDCS (Fig. 3B). There were no whole-brain effects of tDCS or Group when subjects performed the Flanker or Faces tasks.

Figure 3: N-back Task, Trait-Smoking Model (Whole brain results).

(A) 9 regions affected by tDCS during high WM-load (3–0 contrast), whole brain corrected FWE α < 0.01 (p-voxel < 0.001, cluster threshold > 85). (B) Graphical representation of numerical results, by tDCS condition, in each region; for pattern only. L = Left, R = Right, Ph = Parahippocampal Gyrus, Hip = Hippocampus, Pcn = Precuneus, MCC = Middle Cingulate Cortex, IPL = Inferior Parietal Lobule, SMG = Supramarginal Gyrus, STG = Superior Temporal (Heschl’s) Gyrus, pIns = Posterior Insula, MTG = Middle Temporal Gyrus, IFG = Inferior Frontal Gyrus (p. Orbitalis).

Nicotine-State Model:

There was a tDCS*Patch interaction in six DMN-associated regions during the high WM-load condition, overlapping with the nine regions observed in the Trait-Smoking model (e.g. ACC, hippocampus, PHG, and temporal regions, Fig. 4A, Table S5). An-dlPFC tDCS accentuated the downregulation of these regions to a greater extent in the sated (nicotine patch) than in the withdrawal (placebo patch) condition (Fig. 4B). An-vmPFC tDCS again resembled Sham (Fig. 4B). Finally, the nicotine (vs. placebo) patch appeared to restore diminished DMN suppression, and reduce ‘excessive’ thalamic and caudate activity, associated with 12h nicotine abstinence. Specifically, during the N-back high WM-load condition, there was accentuated deactivation in the precuneus (posterior DMN node) in the nicotine patch condition following 12h nicotine abstinence, compared to placebo patch (Fig. S9). During the Flanker task, under the nicotine patch condition, there was reduced activity in the bilateral thalamus and caudate, compared to the placebo patch condition (Fig. S10). There were no tDCS effects in the Nicotine-State models of Flanker or Faces, and no effects of Patch on Faces.

Figure 4: N-back Task, Nicotine-State Model (Whole brain results).

(A) 6 regions displaying a tDCS * Patch interaction during high WM-load (3–0 contrast), whole brain corrected FWE α < 0.01 (p-voxel < 0.001, cluster threshold > 87). (B) Graphical representation of numerical results by tDCS condition, in each region, separated by patch condition (nicotine patch, top; placebo patch, bottom); representation for pattern only. L = Left, R = Right, ACC = Anterior Cingulate Cortex, MTG = Medial Temporal Gyrus, Ph = Parahippocampal Gyrus, Hip = Hippocampus, SMG = Supramarginal Gyrus.

Brain-Behavior Relationships:

tDCS did not modify brain-behavior relationships in the tasks, for either model.

tDCS outcomes

tDCS was well tolerated and effectively blinded (Fig. S11). E-field modeling showed that active tDCS produced a generalized, pre-frontal maximum E-field, which extended somewhat along the cortical midline (Fig. 1D). The normal component of the E-field for An-dlPFC tDCS (Fig. 1C, left) showed an inward field moderately lateralized over L-dlPFC, and an outward field over R-vmPFC; as expected, current flow reversed for An-vmPFC (Fig. 1C, right).

Discussion

We tested whether single-session tDCS of the L-dlPFC/R-vmPFC could acutely modify large-scale brain networks and cognitive task performance implicated in the NWS in a cohort of nontreatment-seeking smokers and matched nonsmokers. An-dlPFC tDCS enhanced DMN suppression during a WM task, and SN activation (ACC), during an error monitoring task (Table 2). Smokers showed a greater effect of tDCS within the ACC than nonsmokers. While both groups were sensitive to the DMN-related tDCS effects, smokers responded to tDCS more in the nicotine sated than the deprived state. Given that failed DMN suppression correlates with NWS severity (43) and predicts smoking relapse (67), it is salient that tDCS enhanced DMN suppression during high WM-load. Although tDCS did not affect task-related behavioral measures, nicotine satiety alleviated withdrawal-induced impairment in attention (omissions) and consequent task engagement (accuracy, response speed), and enhanced deactivation of the precuneus (a key posterior DMN node) following 12h nicotine abstinence. Together, these data suggest that the NWS circuit dysregulation, partially alleviated by nicotine replacement, may be further improved by “adjuvant” anodal-(L)dlPFC/cathodal-(R)vmPFC tDCS.

Table 2:

Summary of Key Findings for An-dlPFC tDCS

Network/Region	Predicted An-dlPFC tDCS effect	Cognitive Task	Task Contrast	Analysis	Model	An-dlPFC tDCS Result	Corrected p-value
Salience Network	Enhance	Flanker	Conflict None, Med, High	ROI (Smk-Trait)	Task Difficulty * tDCS * Group	Enhanced dACC activity in smokers more than nonsmokers	0.03
				Whole brain (Smk-Trait)	Task Difficulty * tDCS * Group	No effect observed	.
Executive Control Network	Enhance	N-back	(3–0) back	ROI (Smk-Trait/Nic-State)	tDCS * Group; tDCS * Patch	No effect observed	-
Default Mode	Reduce	N-back	(3–0) back	Whole brain (Smk-Trait)	tDCS * Group	Enhanced reduction of DMN-related activity in response to high WM-load	<0.01
Network				Whole brain (Nic-State)	tDCS * Patch	In smokers, induced DMN suppression more in sated condition	<0.01
Amygdala (prefrontal effector)	Reduce	Amygdala	(Faces - Shapes)	ROI (Smk-Trait/Nic-State)	tDCS * Group; tDCS * Patch	No effect observed	-

An-dlPFC affected widespread DMN and SN regions, including medial-temporal DMN nodes and ACC. Notably, the affected regions were largely distal from the electrode sites, a finding consistent with other tDCS-MRI studies (For review, 1). We consider two possible mechanisms. First, the network organization of the brain, as observed in resting state analysis (20–22,83), may explain distal effects. Modulating one node in a network likely alters related nodes within or between networks (84). Specific to the anticorrelated ECN–DMN relationship (20,85), we predicted and observed that anodal-dlPFC stimulation would downregulate the DMN. Our result parallels the literature: excitatory, high-frequency repetitive TMS delivered to the L-dlPFC in depressed patients downregulates the subgenual cingulate and improves depressive symptoms (86,87), and dlPFC tDCS in methamphetamine abusers downregulates the DMN and reduces craving (88,89). Further, because anodal-dlPFC/cathodal-vmPFC stimulation are paired in our An-dlPFC condition, inhibitory vmPFC stimulation may have also directly contributed to DMN suppression.

Distal activations can be further explained by E-field modeling. The L-dlPFC/R-vmPFC montage produced high E-field strengths in prefrontal and anterior midline structures. CSF-mediated current flow (90) may have contributed to ACC activation during prefrontal tDCS. Further, the E-field normal component validates that An-dlPFC generally produced inward current over L-dlPFC and outward current at R-vmPFC (Fig. 1C, left), and thus could have accessed the ECN and DMN networks through these nodes. Current over midline dorsal-medial PFC (dorsal-medial DMN sub-system component (30,91)) may have additionally modulated DMN. Taken together, our findings support our hypothesis (Fig. S1) that tDCS may partially alleviate large-scale network dysfunction associated with the NWS.

It is notable that sated-smokers responded to tDCS in a similar fashion as the nonsmoker group, possibly because tDCS may act through a common physiological mechanism regardless of disease trait (i.e. DMN suppression). While trait did not affect outcomes in this sample, we observed an influence of nicotine state within smokers. An-dlPFC most affected DMN in smokers during satiety, compared to during withdrawal. This may be due to reduced cortical flexibility or plasticity during withdrawal, further supported by the impaired behavioral task engagement. Similarly, it has been reported that nicotine deprivation diminishes, and acute nicotine restores, facilitatory plasticity in smokers (92). Withdrawal-related reduced plasticity may be secondary to nicotinic acetylcholine receptor (nAChR) desensitization with chronic smoking (93–95), necessitating acute nicotine to restore synaptic calcium dynamics (96,97), although this is speculative as the current data cannot speak to receptor dynamics. Nevertheless, our data do suggest that tDCS may be most clinically efficacious when combined with nAChR agonism.

We originally predicted that An-vmPFC (i.e. cathodal-dlPFC) would have an inverse effect on network outcomes, such that it would exaggerate NWS-related circuit dysregulation; however, in our hands, An-vmPFC generally resembled sham, or had inconsistent directionality with weak effect size. While we expected cathodal-dlPFC stimulation would be inhibitory to ECN, consistent with motor cortex findings (98), the lack of effect corroborates previous MR-tDCS reports in which cathodal and anodal tDCS did not induce opposing BOLD activity (53,54). tDCS-induced excitability – an area of active research – is affected by neuronal morphology (99,100), non-neuronal cellular neighbors (101,102), and contemporaneous brain activity (103). The lack of anodal-vmPFC effects may also be attributable to regional anatomy.

DMN modulation by tDCS is physiologically and clinically relevant to the NWS. An NWS circuit model (34, 35) describes two components: hyper-excitable, interoceptively-oriented SN-DMN & within-DMN circuits, and under-responsive, exogenously-oriented SN-ECN & within-ECN circuits. Support for DMN and SN-DMN deficits is well documented (36,44,104–109), although ECN and ECN-SN deficits have been somewhat more elusive. One possibility is that attentional deficits (110,111) are primarily due to exaggerated DMN and SN-DMN activity, promoting a shift of limited attentional capacity toward interoceptive, self-generated, and task-unrelated thoughts (91). ECN or cognitive deficits may represent downstream outcomes of shifted attention, as opposed to underlying processing deficits (112). Self-generated thoughts drive DMN activity (113), but are not necessarily passive or “task-negative”; they include planning, scene construction, and somatosensory awareness, all of which consume ‘executive’ resources like other types of internal attention (114). Dysfunction in self-generated thought may consist of aberrant content (e.g. negative valence thoughts, rumination, or fixation, such as about smoking) or aberrant context (e.g. thoughts occur intrusively during another task) (91). Because self-generated thoughts themselves require attention, fronto-parietal and cingulo-opercular attentional networks may remain active, possibly explaining variable reports of NWS ECN deficits. In support of this interpretation, we found that smokers (deprived and sated) responded appropriately to the N-back WM task with ECN-like activity, but the nicotine-deprived state was associated with hyperactive DMN activity and impaired task performance. An NWS model centered on attentional bias toward DMN activity is also supported by rsFC (44), and parallels findings that DMN hyper-excitability worsens clinically relevant outcomes in addictive disorders (14). Given the above, it is notable that An-dlPFC tDCS broadly downregulated DMN nodes, evidence that tDCS may support DMN-suppression in NWS.

The present data support further exploration of the use of tDCS as a complementary therapy to other, standard treatments for nicotine dependence, especially the dysregulated attentional and cognitive processes in the NWS. Given that SN, DMN and ECN dysfunction is hypothesized to underly a range of neuropsychiatric diseases (26, 115), our finding has broad mechanistic implications for therapeutic application of tDCS to treatment-resistant psychopathology beyond nicotine dependence. Indeed, tDCS has been successfully tested as a complement to standard therapy in depression, in which combined tDCS +antidepressant medication improved outcomes more than either alone (116). A similar finding was observed in bipolar disorder, alongside standard mood stabilizers (117). Within substance use disorders (SUDs), tDCS shows encouraging clinical effects in stimulant (88, 118–120), and alcohol (121–124) dependence. Specific to nicotine, tDCS reduces craving (56, 57, 61), negative affect (125), and cigarette consumption (58–60), and improves outcomes in an NWS animal model (126). Our findings suggest that DMN suppression, and SN enhancement, may be important mechanisms by which tDCS exerts these effects. Together with these reports and our finding that tDCS most affects smokers in the nicotine-sated state, we propose that An-dlPFC tDCS may be most efficacious as “adjuvant” therapy for nicotine cessation, to support adaptive neurophysiological changes (DMN suppression, SN enhancement) synergistically with standard-of-care (e.g. nicotinic replacement, cognitive behavioral therapy).

Finally, our findings replicated previous Trait-Smoking and Nicotine-State characteristics of cigarette smoking. Within smokers (Nicotine-State), the withdrawal condition (placebo patch) impaired task performance, consistent with literature (110,127–129). Nicotine satiety alleviated both the task deficits, as previously reported (43, 130, 131), and enhanced precuneus suppression (DMN node) during WM. nAChR agonism (e.g. nicotine, varenicline) suppresses DMN nodes (105, 132–136), enhances DMN-ECN anti-connectivity (43), and reduces SN-DMN connectivity (37, 44). We observed a withdrawal-associated increase in thalamic activity, which was attenuated by nicotine. The reported effects of nicotine in the thalamus are variable (105, 137–139). Its high density of nAChRs (140), and relationship to SN (27) and attention (141, 142), warrant further investigation of its role in NWS. In terms of Trait-Smoking, sated-smokers and nonsmokers generally performed equally on the tasks, except for modestly better N-back and Faces performance in sated-smokers. This may reflect minimal cognitive enhancing effects of transdermal nicotine, which, while calibrated to individual cigarettes per day, represented a substantial dosage. While acute nicotine administration can improve task performance (143), chronic exposure results in a new allostatic state (12,144) that requires nicotine for normal functioning (110,145). Thus, it is expected that a sated smoker will perform approximately equally to a matched nonsmoker, as we found for other task behaviors.

Limitations

Our single-session design naturally constrains generalization regarding either the duration of observed BOLD changes or how outcomes change with multiple sessions, as would likely be applied therapeutically. Multi-session tDCS influences outcomes in mood disorders (116,117,146), post-stroke aphasia (147) and SUDs (57,59,120,124). The lack of multiple sessions combined with the modest cohort size may have contributed to our null effects on task behavior, which is expected to be less sensitive to stimulation than the underlying neurocircuitry. The acute nature of the intervention (three tDCS conditions within one visit) precluded pre/post-tDCS assessment of NWS clinical correlates (e.g. craving, affect). We further note that the pattern of prefrontal stimulation was somewhat diffuse in this design, a characteristic of the electrode size and configuration. We were unable to assess factors of gender or age in our sample: while both features were frequency-matched across groups, gender was confounded by IQ (p = 0.04) across the total sample, and age followed a skewed distribution with uneven representation of age groups. We encourage future studies to directly assess the influence of these factors on tDCS response. Lastly, as the smoker group was relatively small (n=15), these data should be considered exploratory.

Conclusions

tDCS and other noninvasive brain stimulation techniques are emerging interventions at the crossroads of basic neuroscience research and clinical translation, with the potential to alter brain network connectivity dynamics, either as alternative or adjuvant therapies for various neuropsychiatric diseases. Nicotine dependence, and SUDs in general, are prime examples of the failure of current pharmacotherapies to elicit a satisfactory treatment response (9,148). We present the first evidence that tDCS may have a functional neurophysiological effect on two core large-scale networks implicated in NWS. We show that single-session An-tDCS to L-dlPFC accentuates DMN downregulation and SN upregulation during task performance, and that these effects are accentuated in nicotine-sated smokers. Given the general contribution of SN, DMN, and ECN dysfunction across neuropsychiatric disorders, our results support a mechanistic rationale for therapeutic development of tDCS in psychiatry.