Thirty-nine regular smokers (age 24–55 years) participated in the current study. The details of subject recruitment have been described in our prior work [40] and are provided in the Supplementary material. The study was conducted according to a protocol approved by the Institutional Review Board of Yale University (approval number: 2000022986). Written informed consent was obtained from each individual prior to the study. A final sample size of 28 (13 women) was included in the data analyses, after excluding participants who did not complete scanning (n = 7), smoked during the required abstinence (n = 2), or had excessive head motion (n = 2).

Participants completed one fMRI session with overnight abstinence from smoking and the other session with ad libitum smoking. The two sessions for each participant were 1–2 weeks apart and counter-balanced across subjects. Exhaled breath carbon monoxide (CO) level was tested for verification of overnight abstinence. Urine toxicology tests were performed prior to each scan for illicit substances and quantifying the cotinine levels. Before the first magnetic resonance (MR) scan, participants reported daily consumption of cigarettes, years of smoking, and nicotine addiction severity using the Fagerström Test for Nicotine Dependence (FTND; range 0–10). We computed pack years for each participant by multiplying the number of packs of cigarettes smoked per day by the number of years of smoking to reflect lifetime exposure to tobacco. We also assessed novelty seeking, harm avoidance, and reward dependence with the TPQ [5].

Prior to each scan, we recorded the time since last cigarette in hours for each participant (i.e., abstinence duration). Twenty-one participants (13 women) were assessed for withdrawal symptoms with reference to the past 24 hours using Hughes-Hatsukami Withdrawal Questionnaire [41], craving to smoke with a brief, 10-item version of the Questionnaire of Smoking Urges [42], and state anxiety with State-Trait Anxiety Inventory [43].

We examined sex differences in the demographic and clinical measures with two-sample t tests. The 2 (abstinence vs. satiety) by 2 (men vs. women) mixed-model analyses of variance (ANOVAs) were performed on the CO level, withdrawal duration, withdrawal symptom severity, craving to smoke, and state anxiety. Given that the urine cotinine level was an ordinal variable, we used a nonparametric test of mixed-model ANOVA instead. We computed partial correlation coefficients between TPQ trait scores and FTND score and clinical characteristics that showed significant state differences with age as a covariate in men and women combined and separately. We used slope tests to examine the sex differences in the correlations [44, 45, 46].

Imaging was conducted with 3.0 Tesla Prisma whole-body magnetic resonance imaging (MRI) system and 64-channel head coil (Siemens Medical Solutions, Inc., Erlangen, Germany), with scout scans, high-resolution MPRAGE, and BOLD scans, as in our published work [47]. The details of MRI data acquisition are provided in the Supplement. We preprocessed the resting state fMRI data as described earlier [40, 48], including realignment, co-registration, normalization, smoothing, regression of confounding signals (i.e., white matter, cerebrospinal fluid, whole-brain mean, and physiological signals), temporal band-pass filtering, and scrubbing. On average, 18.11% and 18.79% of the time points were removed across subjects for abstinence and satiety scan, respectively. There was no significant difference in the numbers of scrubbed time points between abstinence and satiety scans (t = 0.19, p = 0.848).

As with our previous studies [45, 48], the bilateral VS mask was generated by cytoarchitectonic and topographical criteria [49], and the masks of bilateral DS as well as subregions of the DS, including bilateral caudate, putamen, and pallidum were obtained from the Automated Anatomic Labeling atlas [50]. We used these masks as seed regions (Fig. 1A) to compute whole-brain rsFC. We extracted the $\beta$ estimates of the regions of interest (ROIs) revealed in whole-brain rsFC analyses for each state and computed the $\beta$ differences between states for linear correlations with FTND score, TPQ traits, and clinical measures that showed significant differences between states. We performed the analyses in men and women combined and separately.

Fig. 1.

Striatal masks and a cluster identified in whole-brain analysis. (A) Masks of ventral striatum (VS), caudate, putamen, and globus palliums (GP). (B) A cluster of left insula extending to posterior OFC and IFGpo showed stronger rsFC with VS in abstinence vs. satiety. Color bar shows T value and Cohen’s d for the significant cluster. OFC, orbitofrontal cortex; IFGpo, inferior frontal gyrus pars orbitalis.

The mean and standard deviation (SD) values of demographic and clinical measures are presented in Table 1. Two-sample t tests showed no significant sex differences in any of the demographic or clinical metrics (p’s $\ge$ 0.124). The 2 (state: abstinence and satiety) by 2 (sex: men and women) mixed-model ANOVAs showed a significant interaction effect on the CO level (F = 9.91, p = 0.005) but not on any other clinical measures (p’s $\ge$ 0.753). Simple-effect tests showed that the CO level was significantly lower in abstinence vs. satiety in women (p 0.001) but not in men (p = 0.060) and there were no sex differences in the CO level in either abstinence (p = 0.371) or satiety (p = 0.296). The nonparametric interaction test for mixed-model ANOVA showed no state-by-sex interaction effect on urine cotinine level (F = 0.10, p = 0.753). The main effects of state were significant on the CO level, withdrawal interval, withdrawal symptom severity, and craving to smoking (p’s $\le$ 0.002), but not on the cotinine level (p = 0.753) or state anxiety (p = 0.060). None of these clinical measures showed significant main effects of sex (p’s $\ge$ 0.148).

Across all participants, the TPQ NS, HA, or RD scores were not significantly correlated with FTND score (p’s $\ge$ 0.498), or any clinical measures in abstinence (p’s $\ge$ 0.113) or satiety (p’s $\ge$ 0.078). In men alone, HA score was positively correlated with withdrawal symptom severity in abstinence (r = 0.82, p = 0.025) but not in satiety (r = 0.31, p = 0.501), though slope test showed no differences in the correlation (t = 0.33, p = 0.746). In contrast, women did not show a significant correlation between HA score and withdrawal symptom severity in abstinence (r = 0.27, p = 0.453) or in satiety (r = 0.42, p = 0.223). The correlation of HA score and withdrawal symptom severity during abstinence was significantly stronger in men than in women (slope test, t = 2.31, p = 0.036). RD and FTND scores were negatively correlated in women (r = –0.65, p = 0.040) but positively correlated in men (r = 0.45, p = 0.312) and the slope test showed significant differences in the correlation (t = 2.16, p = 0.042). None of the other correlations were significant in men and women examined together or separately.

In whole-brain analyses of the full sample, paired-sample t test of state did not show any clusters with significant differences in rsFC with bilateral DS, caudate, putamen, or pallidum. With bilateral VS as a seed, the rsFC with a cluster in the left anterior insula (AI) extending to posterior orbitofrontal cortex (pOFC) and inferior frontal gyrus, pars orbitalis (IFGpo) (cluster size k = 74, MNI coordinates x = –34, y = 22, z = –12, peak Z = 4.38) was stronger in abstinence vs. satiety (Fig. 1B). A mixed-model sex $\times$ state ANOVA of the $\beta$ showed a stronger VS-insula rsFC in abstinence vs. satiety [F(1, 26) = 36.12, p 0.001], as expected, but no significant sex main [F(1, 26) = 3.54, p = 0.071] or interaction [F(1, 26) = 0.87, p = 0.359] effect.

Whole-brain paired-sample t tests were also performed for men and women separately. However, no clusters showed significant rsFC with bilateral VS, DS, caudate, putamen or pallidum in either men or women at the same threshold.

With age as a covariate, the difference of VS-insula rsFC between abstinence and satiety was negatively correlated with HA score in men and women combined (r = –0.54, p = 0.021). This correlation was marginally significant (r = –0.74, p = 0.055) in men alone but not significant in women alone (r = –0.43, p = 0.218), and the slope test showed no significant sex difference in the correlations (t = 0.10, p = 0.918). The difference of VS-insula rsFC between states was not significantly correlated with NS, RD, or FTND score or with the differences of withdrawal symptom severity or craving to smoke between states in all (p’s $\ge$ 0.190), men (p’s $\ge$ 0.138), and women (p’s $\ge$ 0.232) alone, with age as a covariate.

We also performed regressions between VS-insula rsFC and TPQ subscores and clinical measures separately for satiety and abstinence, and the results are shown in Supplementary Table 1. Notable were positive correlations of VS-insula rsFC with FTND scores though the correlation was significant during satiety (r = 0.570, p = 0.014) but not abstinence (r = 0.346, p = 0.160).

We observed stronger VS rsFC with AI extending to pOFC and IFGpo in overnight abstinence vs. satiety, in line with an earlier study associating 24-h nicotine abstinence with higher VS rsFC with the insula, superior temporal gyrus and anterior/mid-cingulate cortex [39]. The latter study also reported decreased DS rsFC with medial prefrontal cortex, posterior cingulate cortex, hippocampus, and supplementary motor area over 24-h abstinence vs. satiety; in contrast, we did not find any significant differences in the DS connectivity in overnight abstinence vs. satiety. These findings collectively suggest altered VS-insula rsFC at the very early stage of abstinence, which may be followed by changes in DS rsFC. Of note, Sweitzer, Geier [39] demonstrated elevated VS rsFC in non-relapsers but the opposite in relapsers at 3 weeks of follow-up and the same patterns of changes in DS rsFC across both groups, associating altered VS but not DS connectivity during early abstinence with smoking cessation outcomes.

Neurobiological models of addition implicated both striatum and insula in the addictive processes of binging/intoxication, withdrawal/negative affect, and craving [51, 52]. Previous research demonstrated lower VS-insula connectivity in non-treatment-seeking cocaine users as compared with healthy controls [53]. Here, we observed that overnight abstinence induced higher VS connectivity with the insula extending to pOFC and IFGpo, implicating the striato-insula and corticostriatal circuits during acute withdrawal. The striato-insula circuit underlies motivation and reward processing, integration of affective and cognitive function, as well as action selection and planning [54, 55]. The striato-insula circuit may support the motivation to use drugs as a result of aversive bodily states in abstinence [56]. Unexpectedly, we did not find any significant correlations between changes in VS-insula rsFC and withdrawal symptom severity or craving to smoke, likely because of the small sample size and/or relatively low levels of nicotine dependence of the participants. Note that the VS-insula rsFCs were positively correlated with FTND scores (Supplementary Table 1), suggesting VS-insula rsFC as a feature of smoking problem. Thus, if verified, those with more severe smoking problem would show higher VS-insula rsFC and VS-insula rsFC would become even stronger during withdrawal.

We found higher levels of HA trait in association with more severe withdrawal symptoms during overnight abstinence in men but not in women, with sex differences confirmed by the slope test. An animal study found that nicotine withdrawal was accompanied by a significant and persistent loss of striatal serotonergic activity, starting from initial abstinence [57]. The HA trait, as characterized by excessive worrying, pessimism, shyness, fearfulness, doubtfulness, and easy fatigability is related to low serotonergic activity [58, 59]. We speculate that striatal serotonergic activity may play a role in supporting sex differences in the relationships between HA trait and withdrawal symptom severity, an issue that can be investigated with molecular imaging in humans as well as imaging and electrophysiology in rodent models. Our findings are also largely in line with prior evidence demonstrating that HA predicted 12-h nicotine abstinence-induced increases in negative affect and urges to smoke [24]. We also observed that RD was negatively correlated with FTND scores in women, in accord with social detachment and insensitivity to social approval as a risk factor of substance misuse [7, 60]. In contrast, men showed positive though non-significant correlation between RD and FTND score, consistent with a study associating RD with craving in response to both smoking and stress vs. neutral cues [61]. Our findings suggest sex as a critical factor to consider in studies of the influences of RD traits on substance use and misuse.

Prior research has demonstrated sex differences in resting state network connectivities in smokers. For instance, women vs. men showed stronger functional coupling of the hippocampus/amygdala with bilateral AI, rostral anterior cingulate cortex, and inferior parietal lobule [62] and greater connectivity within the default mode network but no significant differences in the connectivity within the reward (i.e., striatum and orbitofrontal cortex) network [63]. Although we did not observe that men and women differed in VS whole-brain connectivity, we showed that higher HA was associated with less abstinence-induced changes in VS-insula rsFC, a relationship that appeared driven by males, consistent with earlier findings that people with higher HA showed lower regional rsFCs and insular-opercular network efficiency during resting [64]. Both the VS [65] and insula [66] showed higher activity during risk-taking decisions in positive correlation with HA. Moreover, the potential sex difference indicates that sex may modulate the relationship between HA trait and abstinence-related changes in VS-insula rsFC. In our sample, men smoked more heavily and had greater pack years as compared to women. A larger sample with men and women well-matched in nicotine use metrics may help in investigating this issue further. On the other hand, male smokers higher in HA experienced more severe withdrawal symptoms and across subjects VS-insula rsFC was positively correlated with FTND score at baseline. Thus, the findings suggest that VS-insula rsFC may reflect a relatively stable marker of individuals high in HA—likely in relation to the psychological processes of risk taking—that is not impacted by acute withdrawal from smoking.

We also observed higher levels of RD trait in association with less severe nicotine dependence in women and potentially the opposite in men. A meta-analysis on sex differences in TPQ traits in healthy individuals found that women scored higher in RD than men, reflecting stronger female endorsement of social desirability [67]. Furthermore, women with lower RD scores may show difficulties in social attachment and greater susceptibility to depression or adjustment problems [68, 69, 70], which in turn is associated with nicotine misuse. Together with our finding that speaks to the potentially contrasting, sex-specific roles of RD in substance misuse, men may engage in drinking and smoking to seek social approval and interaction whereas women do so because of social distancing and isolation. More research is needed to explore this hypothesis.

The current study has several limitations. First, our findings are of a small sample size and preliminary. In particular, the smokers showed low-to-moderate levels of dependence, and it remains unclear whether the findings would replicate in heavier smokers. Secondly, male vs. female smokers showed higher pack years and more (though not significantly different) severe withdrawal symptoms in our study. A better-matched cohort of men and women may help in confirming the findings. Finally, changes in striatal connectivity may be associated with motivation to quit and other psychological processes involved in chronic smoking. Longitudinal studies and more thorough clinical and behavioral assessments are needed to expand on the findings.