2.1.

Participants

Thirteen HND were recruited from a local drug rehabilitation center in Macau SAR. Seven active nicotine-dependent (ND) subjects matched with the HNDs (i.e., age, education level, years of smoking, and daily cigarette consumption) were enrolled from the local community. The HND and ND were diagnosed with substance use disorders (SUDs) (i.e., OUD and tobacco use disorder (TUD), respectively) based on the diagnostic and statistical manual of mental disorders - 4th edition (DSM-IV) diagnostic criteria by clinicians. HND were enrolled based on their abstinence for at least 3 months in order to eliminate the effects of acute heroin withdrawal. Total 110 healthy control subjects (C) were recruited from the University of Macau campus via online and social media advertisement. All subjects participating in the neuropsychological assessment were urine-drug screened. Seven HND were able to proceed with the study, and thus, 7 matched C and ND were selected independently for the neuroimaging recording. The inclusion criteria of all subjects were aged between 18 and 65 years, right-handed, Cantonese speakers, and had normal or corrected-to-normal vision. The exclusion criteria were a history of neurological illness, brain surgery, other psychiatric conditions, or reported use of psychoactive substances (except methadone and nicotine) at least 72 h before assessment. Total 21 matched and included subjects were proceeded to neuroimaging assessment. The enrollment scheme of this study is shown in Fig. 1. The study was conducted in accordance with the Declaration of Helsinki of 1975. The protocol was approved by the Medical Ethics Committee of the University of Macau. All subjects provided written informed consent before participating and were compensated upon completion of the experiment.

Fig. 1

Scheme of study enrollment, grouping, and evaluable participants for neuropsychological and neuroimaging measurements. The inclusion criteria of all subjects were between 18 and 65 years of age, right-handed, Cantonese speakers, and had normal or corrected-to-normal vision. The exclusion criteria were a history of neurological illness, brain surgery, other psychiatric conditions, or reported use of psychoactive substances (except methadone and nicotine) at least 72 h before assessment. A nonclinical sample of 110 healthy subjects was enrolled to ensure each stimulus in cross-cultural task reached criterion level of consensus. Total 21 matched and included subjects were proceeded to neuroimaging assessment. The demography of the enrolled subjects is in Table 1.

Additional details regarding the methods and materials in this study are available in Appendix A. It includes supplementary information about the medications status of the participants, the image preprocessing and its analysis, and the FC network analyses mentioned in the main text below.

2.2.

Neuropsychological Assessment

The neuropsychological battery in the study was designed to assess the EF, ToM, and behavioral stressors of the subjects. The neuropsychological domains, functions, and test administrated are tabled in Appendix B (Table 7).

EF refers to a set of abilities to carry out goal-directed behaviors, which were assessed in five domains in this study: inhibition, shifting, updating, access, and processing speed (PS). The administrated tests include stroop (STP),⁹ block design (BD),¹⁰ WM,^10,11 verbal fluency (VF),¹² and PS¹⁰ tasks.

ToM refers to the ability to perceive and attribute mental states, including intentions and emotions to both self and others³¹ and social intelligence,³² and it was assessed in the emotion-recognition domain by the Reading the Mind in the Eyes (RME) test.³¹ RME is an advanced ToM test that involves mapping mental-state lexicon and its semantics to fragments of facial expressions of mental states of others (i.e., only the part of the face around the eyes of another person). Considering cross-culture variation, the Asian RME task, which was adopted from Adams et al.,³³ was translated into Chinese from Japanese, and the Chinese version of the ToM test was piloted on a nonclinical sample of 110 healthy subjects to ensure each stimulus reached criterion level of consensus. The RME scores and the averaged reaction time of the participants are measured.

The stressors were assessed in four domains: childhood stress, depression, anxiety, and hyperactivity. The neuropsychological tests for assessing stressors include adverse childhood experiences (ACE) questionnaire,¹⁴ Beck depression inventory-II (BDI-II),¹⁵ state-trait anxiety inventory (STAI),¹⁶ and adult attention deficit/hyperactivity disorder self-report (ASRS-v1.1) symptom checklist.¹⁷ All enrolled subjects were required to complete a neuropsychological assessment performed by a trained psychological counselor.

2.3.

Image Acquisition

All fNIRS examinations were conducted on a continuous-wave (CW) instrument (CW fNIRS system; TechEn Inc. Milford, Massachusetts) with 4 sources and 8 detectors, generating 14 channels. The regions of interest (ROIs) included the bilateral OFC, dlPFC, and mPFC. The configuration of fNIRS sources and detectors covering the PFC is shown in Fig. 2. Two CW lights at wavelengths of 690 and 830 nm were emitted during recording to detect the neurophysiological hemodynamic signals, that is, the changes in oxyhemoglobin (HbO) and in deoxyhemoglobin (Hb) in blood vessels³⁶ over the PFC. The sampling rate was 50 Hz. A three-dimensional (3-D) digitizer (Polhemus Inc., Vermont) was used to complete the spatial registration of the channel locations, and thus, the Montreal Neurological Institute (MNI) coordinates were obtained (Appendix C, Table 8).

Fig. 2

Summary of 6 ROIs and their associated channels with the arrangement of the 14 channels covering the prefrontal regions. Each source and detector was 3 cm apart: (a)–(c) were generated using NIRS-SPM software³⁴ and (d) generated using AtlasViewer software from Homer2.³⁵

Subjects were instructed to sit comfortably in a quiet dim room and move as little as possible during recordings. An 11-min, eye-closed, resting-state recording was followed by an $\sim 11\text{-}\min$, eye-open, Asian RME ToM task adopted from Adams et al.³³ Resting-state and RME tasks were coded using E-prime 2.0 software (Psychology Software Tools, Inc., Pennsylvania). Before the RME task began, subjects were instructed to place their four fingers: index, middle, ring, and little fingers, on a four-button keyboard numbered from 1 to 4, respectively. Subjects were told that a pair of eyes in a picture would be displayed at the center of the screen for 3 s. During the 3 s, the subject’s task was to try best to think about the emotion of the protagonist in the picture. At the fourth second, four-numbered labels of emotion were then displayed at each corner of the picture while the picture remaining at the center, and the subject’s task was to choose a label that best described the emotion of the protagonist as fast and accurately as they could by pressing a corresponding button on the keyboard. Once the subject pressed the button, a fixation would be displayed for 15 s until the next stimulus was started; total 36 stimuli. The response time (RME_RT) of each stimulus was recorded in microseconds for analysis. An example of an RME stimulus is in Fig. 3.

Fig. 3

An example stimulus of the Asian RME for the fNIRS recording. In 1 trial, an RME stimulus was displayed at the center of the screen for 3 s when a subject’s task was to read the picture and try to think about the emotion of the protagonist. Four numbered adjectives of emotion were then displayed at each corner of the picture at the fourth second, in which the subject’s task was to choose an adjective that best describes the emotion at subject’s fastest effort. The goal of the RME test was to assess the subject’s ability to mentalize feeling and thought of others. The Asian RME was adopted from Adams,³³ and the labels were translated in Chinese as the figure shown from Japanese. Language in English, Chinese, and Japanese as follow: (1) embarrassed, $尷尬$, $きまりが悪い$, (2) guilty, $內疚$, $気がとがめている$, (3) fantasizing: $幻想$, $空想にふけっている$, and (4) concerned (target), $擔心$, $心配そうにしている$. Reproduced with permission from Adams.³³

Nodes and edges are fundamental elements of a network. In this study, nodes were defined by channels, whereas edges were defined as FC among nodes. The fNIRS-based image preprocessing for the FC generation was previously discussed in our work.³⁷ The HbO data were analyzed in the study due to its high sensitivity. For the task data, the time course during the 3-s mental recording and its 17-s hemodynamic delay were extracted for analysis. Whole-brain analysis was used to test the FC strength. A mean group-level correlation matrix of each set data was obtained from the resting-state and ToM-task fNIRS data separately for each study group. The foremost 10% of connections calculated from the population (method details in Niu et al.³⁸) were used as the threshold. Correlations coefficients greater than the predetermined threshold value were considered as edge. The details of the method are available in Appendix A.

2.4.

Statistics

Statistical analyses were performed using SPSS (IBM SPSS Statistics, Armonk, New York). Group differences in demographic, behavioral, and ROI data were investigated using one-way ANOVA and Bonferroni correction. Using a current alternative strategy,^26,39 the matched sample size (i.e., $n_{\text{root}}$) was appropriately adapted to the cost concerns of the study. Prior to any statistical analysis, all correlation values were converted to Fisher $z$-values using the Fisher $z$-transformation, which has been widely used when the sample size is $<30$. Within-group correlations among neuropsychological measures were tested using Spearman’s rank-order test. Linear regression analysis was conducted to examine any link between behavioral measures and FC patterns for each ROI. Multiple regression was performed for all significant results to account for potential differences in age, education level, and IQ. To address the issue of multiple comparisons, Bonferroni correction was applied to the post hoc tests for behavioral and imaging measures.

3.1.

Demographics

Subjects’ demographic characteristics and drug-taking behaviors are summarized in Table 1. Age, education, and sex were significant in the controls because most of them were recruited from our campus, but they did not differ ($p>0.05$) between groups after matching. Seven controls and 7 ND were matched with 7 HND subjects included in the neuroimaging recording. The attempt-to-commit-suicide rates during lifetime in the dependent groups were significantly higher ($p<0.0001$) than the controls. The smoking habitual behaviors, such as daily dose and years of smoking, did not differ between the dependent groups, but the onset age did ($p<0.001$). HND had a mean (SD) smoking onset at age 13.7 (2.0), whereas ND had at 17.0 (1.5).

Table 1

Demographic characteristics for the studied groups.

3.2.

Heroin Plus Nicotine Dependence Enhances Severity of Stressors

Both HND and ND demonstrated significantly ($p<0.0001$) higher levels of anxiety and childhood adversity compared with the controls (Table 2). The depressive- and attention deficit/hyperactivity disorder (ADHD)-like responses were observed only in HND ($p<0.001$). The post hoc tests showed that HND had a significant effect ($p<0.01$) on the severity of hyperactivity, anxiety, and depression. Within the clinical measures (Fig. 4), the depression-like response predicted the level of anxiety in ND ($R^2=0.594$, $F_{\mathrm{1,6}}=7.31$, $p=0.043$, $\beta =1.59$) and hyperactivity ($R^2=0.743$, $F_{\mathrm{1,6}}=14.4$, $p=0.013$, $\beta =0.315$); whereas these variables were only correlated in HND ($r=0.550$; $r=0.118$, $p>0.05$, respectively).

Table 2

Differences between studied groups in EF, ToM, and clinical measures.

Fig. 4

Correlation matrices between neuropsychological variables: EF, ToM, and clinical characteristics. Statistically significant correlations within-group comparisons: (a) healthy controls (C), (b) ND group, and (c) HND group. Neuropsychological tests include STP, BD, WM, VF, and PS in the EF domain, and RME test and its response time (RME_RT) in the ToM domain. Clinical measurements include ACE, BDI, STAI, and the adult ADHD tests. Color bars represent the $Z$ to $R$ correlation coefficient values. Pearson’s correlations were used. White cube represents significant correlations ($p<0.05$) in regression. Black region represents nonsignificance in regression. The neuropsychological tests administrated in this study are detailed in Appendix B (Table 7).

3.3.

Relations between EF and ToM Mechanisms

The EF and ToM performances are shown in Table 2. Both HND and ND showed significantly ($p<0.0001$) poor performance in the STP test (i.e., inhibition domain) compared with the controls. While compared with ND, HND patients showed significantly ($p=0.021$) better performance in STP. There were significant effects of heroin dependence on the BD, and PS ($F_{\mathrm{2,127}}=14.3$, $p<0.0001$, and $F_{\mathrm{2,127}}=5.06$, $p<0.0001$, respectively), indicating chronic heroin consumption mediated and reduced both aspects of shifting and PS skills. A significant main effect of drug dependence ($F_{\mathrm{2,127}}=4.56$, $p=0.012$) indicated that chronic drug consumption impaired the access domain but post hoc comparison of the ND and HND with the controls did not reach significance. No significant effects of heroin and nicotine on the updating domain were found. The longest response time of the RME task in ND compared with HND and controls was observed.

Correlation and further regression analyses were performed. The relationships between EF, ToM, and clinical characteristics of each group are shown in Fig. 4. Within the EF measures in HND, PS was significantly associated with BD (i.e., the shifting domain; $R^2=0.990$, $F_{\mathrm{1,12}}=\mathrm{40,529}$, $p<0.0001$, $\beta =4.06$) and WM (i.e., the updating domain; $R^2=0.384$, $F_{\mathrm{1,12}}=5.62$, $p=0.042$, $\beta =0.616$), which was also linked to BD ($R^2=0.377$, $F_{\mathrm{1,12}}=5.54$, $p=0.044$, $\beta =2.48$). Similar to HND, ND showed a strong association between PS and the shifting domains ($R^2=0.998$, $F_{\mathrm{1,7}}=2699$, $p<0.0001$, $\beta =0.999$). No significant association within the EF measures was observed in controls.

Between the EF and ToM measures, the performance of RME was associated with that of STP ($R^2=0.379$, $F_{\mathrm{1,12}}=5.50$, $p=0.044$, $\beta =2.45$) in HND. There was an association between VF (i.e., the access domain) and the response time of RME ($R^2=0.047$, $F_{\mathrm{1,109}}=5.38$, $p=0.022$, $\beta =-0.001$) in controls. No significant EF–ToM association was observed in ND.

3.4.

Functional Connectivity

Total 21 subjects were matched with age ($F_{\mathrm{2,18}}=1.46$, $p=0.258$), education level ($F_{\mathrm{2,18}}=2.79$, $p=0.088$), and estimated IQ ($F_{\mathrm{2,18}}=0.361$, $p=0.702$) and were included in the neuroimaging analysis.

The resting-state correlation coefficient threshold value was 0.87 ($\text{mean}=0.52$), and the task-induced correlation coefficient threshold was 0.85 ($\text{mean}=0.52$). Group-level HbO-based connectivity maps during resting state and task are shown in Fig. 5. The resting-state functional connectivity (rsFC) map for HND showed consistent results from prior studies.^37,40,41 Consistent with our a priori hypothesis of a positive relationship between clinical stressors and OFC during resting state,³⁷ we discovered enhanced rsFC connectivity strength in the OFC and mPFC but decreased in dlPFC. The HbO-based nodal degree strength of each ROI from resting-state and ToM-task data is available in Tables 3 and 4, respectively, provided with post hoc analyses, whereas the Hb-based data are in Tables 5 and 6, respectively.

Fig. 5

Whole-brain correlation analysis for comparison of brain networks in healthy controls (C), ND, and HND groups. The top column indicates resting-state networks. The below indicates ToM-induced networks. Only the topmost 10% with correlation values greater than the predetermined thresholds ($T_{\text{rest}}$: 0.87 and $T_{\text{task}}$: 0.85) are shown in the figure. The nodes represent the channels. The edges represent the connections. The nodes (red: medial prefrontal regions, blue: orbitofrontal regions, and green: dorsolateral prefrontal regions) are numbered by channel in control group. The weighted nodes are displayed. R: right and L: left. HbO data during resting-state and ToM task are in Tables 3 and 4, respectively, whereas Hb data are in Tables 5 and 6, respectively.

Table 3

Group differences in nodal degree strength in the change of HbO concentration (ROI analysis) during resting state.

Table 4

Group differences in degree strength in the change of HbO concentration (ROI analysis) during RME task.

Table 5

Group differences in nodal degree strength in the change of Hb concentration (ROI analysis) during resting state.

Table 6

Group differences in degree strength in the change of Hb concentration (ROI analysis) during RME task.

In line with evidence that mPFC plays a significant role in ToM cross cultures,^27,33 the ToM-induced connectivity maps for the healthy subjects also demonstrated similar connections within the mPFC and between the mPFC and the dlPFC regions in the left hemisphere from prior studies.^28,31 During RME, we observed decreased FC across the PFC in HND but enhanced FC in ND, including the connections between the orbital portions of middle and superior frontal gyri within the right OFC regions, between the superior frontal gyri within the left OFC, and between the left superior frontal gyri within the left dlPFC, compared with the controls.

3.5.

Relations between Behavioral Performance and Network Organization

The rsFC local and global network topological properties are plotted as a function of sparsity in Fig. 6. Several correlations and multiple linear regression with post hoc analyses were performed with the neuropsychological performances. We observed a significant association ($R^2=0.756$, $F_{\mathrm{1,6}}=15.5$, $p<0.01$, $\beta =8.82$) between WM performance and nodal degree strength on the right orbital part of superior frontal gyrus (oSFG; i.e., channel 4, $x=22$, $y=72$, and $z=6$) in HND [Fig. 6(f)]. At the same region, a significant association ($R^2=0.779$, $F_{\mathrm{1,6}}=17.7$, $p<0.01$, $\beta =\mathrm{13,685}$) between the RME response time and clustering coefficient was also observed in HND [Fig. 6(g)].

Fig. 6

Brain topological network and its relationship with cognitive functions. (a)–(e) Brain functional network properties as a function of sparsity, $S$, in healthy controls (C; black), ND (blue), and HND (red) groups. (a) Clustering coefficient; significant ranges: $0.11<S<0.15$ ($0.009<p<0.032$), $0.22<S<0.29$ ($0.009<p<0.044$), and $0.38<S<0.60$ ($0.009<p<0.030$). (b) Characteristic path length; nonsignificance. (c) Small-worldness; significant at $S=0.1$ ($p=0.047$). (d) Local efficiency; significant ranges: $0.11<S<0.15$; $0.22<S<0.27$; $0.4<S<0.45$; $0.55<S<0.58$, $0.009<p<0.05$. (e) Global efficiency; significant ranges: $0.24<S<0.25$ ($p=0.033$) and $0.35<S<0.36$ ($p=0.037$). (f)–(g) Regression between network property and cognitive function. (f) WM index in the EF domain versus the nodal network property, nodal degree at the right oSFG ($x=22$, $y=77$, $z=6$; $F_{\mathrm{1,6}}=15.5$, $p=0.013$, $\beta =8.82$). (g) RME task in the ToM versus the global network property, clustering coefficient at the right oSFG ($F_{\mathrm{1,6}}=17.7$, $p<0.01$, $\beta =\mathrm{13,685}$).

4.1.

Limitations and Future Prospectives

There are limitations in this study. This study suffers from relatively small sample sizes, due to insufficient grants and the enrollment of HND subjects that met criteria. Compensating for the limited sample sizes, we focused on controllable features (i.e., HND and ND with reduced variance), increased the size of observed effects (i.e., different parameters among groups), and provided visualization of the data. Although it may be way too far to provide generalization to the population, considering consistent hemodynamic patterns in the frontal area observed in another cohort in our previous work, this study may serve as an early guide for, and encourages, future large-scale translational investigations in various areas including, but not limited to, social cognition, identification of biomarkers, and development for brain state-dependent computation in OUD during recovery.

Like fMRI, fNIRS measures hemodynamic signals and solely relies on the principle of neurovascular transduction mechanism to construe neuronal event (i.e., blood flow). It relies on the ongoing spontaneous event to its applications in neuropharmacology to understand the physiological signals of the cerebral cortex. Despite its higher temporal resolution compared to fMRI, fNIRS can only monitor cortical activities. Thus, future prospective neuroimaging work is needed to explore the subcortical areas associated with the endocannabinoid, opioid, dopamine, and the aversive systems.

It bears mention that applying rsFC in the realm of OUD and linking FC with behaviors are still in their infant stages.^48,49 Prior studies have displayed inconsistent findings as a review⁵⁰ highlighted, and thus, their examinations, including this one, have been exploratory in nature. Both reduced^51,52 and enhanced^37,41,53 FC were reported in the PFC of opioid addicts. There have been only a few OUD studies in social cognition. Indeed, noninvasive neuroimaging measurements of spatiotemporal dynamics of network activity and neuropsychological assessment of cognition and behavior have been limited either by the accessibility of neuroimaging tools or by the consideration of the dynamics within the environment.⁵⁴ The inconsistency is probably due in large part to the associated comorbidity and handedness. Differentiation of comorbid phenotypes in accordance with subject cohorts, which may be a major factor accounting for the contrasting reports, is strongly recommended in future work. Utilizing noninvasive neuroimaging that owns high sensitivity and relatively high temporal and spatial resolutions at affordable cost will be optimal. Multimodal approaches will also be helpful in understanding brain network and its correlate with behavioral measures. It is worth noting that variables of circuit connectivity, thickness of cerebral cortex, and cognitive performance are often age-related; investigators should be cautious of the ages of the enrolled subjects and ensure that they are insignificant between study groups in future neuroimaging studies.