Win and Loss Responses in the Monetary Incentive Delay Task Mediate the Link between Depression and Problem Drinking

Depression and alcohol misuse, frequently comorbid, are associated with altered reward processing. However, no study has examined whether and how the neural markers of reward processing are shared between depression and alcohol misuse. We studied 43 otherwise-healthy drinking adults in a monetary incentive delay task (MIDT) during fMRI. All participants were evaluated with the Alcohol Use Disorders Identification Test (AUDIT) and Beck’s Depression Inventory (BDI-II) to assess the severity of drinking and depression. We performed whole brain regressions against each AUDIT and BDI-II score to investigate the neural correlates and evaluated the findings at a corrected threshold. We performed mediation analyses to examine the inter-relationships between win/loss responses, alcohol misuse, and depression. AUDIT and BDI-II scores were positively correlated across subjects. Alcohol misuse and depression shared win-related activations in frontoparietal regions and parahippocampal gyri (PHG), and right superior temporal gyri (STG), as well as loss-related activations in the right PHG and STG, and midline cerebellum. These regional activities (β’s) completely mediated the correlations between BDI-II and AUDIT scores. The findings suggest shared neural correlates interlinking depression and problem drinking both during win and loss processing and provide evidence for co-morbid etiological processes of depressive and alcohol use disorders.


Introduction
Depression and alcohol misuse are two leading, comorbid, causes of disability [1][2][3][4][5]. Individuals with depression are more likely to drink to cope with negative mood, elevating the risks in developing an alcohol-use disorder (AUD) [3,6]. Drinkers, especially those with AUDs, often experience depression [7], which in turn leads to more drinking [8,9]. Indeed, co-occurring depression and alcohol misuse are known to dispose individuals to greater severity and more frequent relapse of both conditions [10]. In particular, investigators have suggested a causal pathway whereby alcohol dependence increases the risk of major depression rather than vice versa [1]. Specifically, with alcohol use and depression symptom severity quantified at baseline and 1-year follow-up, the structural equation model with AUD leading to major depression showed the best fit [11]. However, it remains unclear whether this is also true of drinkers with mild to moderate alcohol use severity.
Reward delivers pleasure and drives motivated behaviors. Investigators have employed the monetary incentive delay task (MIDT) or card-guessing task to identify the neural responses to win and loss [12,13]. Individuals with depression relative to controls demonstrated altered reward-related activations [14][15][16]. For instance, blunted striatal activity in response to monetary reward and loss has been reported in patients with depression [17][18][19] and anhedonia [20]. Ventral striatal hypoactivity during anticipation of win All participants were evaluated with the Alcohol Use Disorders Identification Te (AUDIT), Beck's Depression Inventory (BDI-II), and the Fagerström Test for Nicotin Dependence (FTND). A 10-item instrument to screen for harmful drinking, the AUDI assesses the frequency of alcohol consumption, dependence, and associated harm. Eac item is scored from 0 to 4 and the total score ranges from 0 to 40, with a higher sco indicating greater severity of problematic alcohol use [41,42]. Five of the forty-thre participants with the highest AUDIT scores (all > 14) also met criteria for alcohol abus The BDI-II is a 21-item assessment of the presence and severity of depression symptom within the prior 2 weeks, with each item scored 0 to 3. A total score of 0 to 13, 14 to 19, 2 to 28, and 29 to 63 indicates minimal, mild, moderate, and severe depression, respective [43][44][45]. Three of the forty-three participants scored > 19 and had moderate severity depression. The FTND assesses the severity of cigarette consumption, compulsion smoke, and physical dependence on nicotine, with a range of 0-10. A higher FTND sco indicates greater severity of nicotine dependence [46]. Each subject completed two 10-m runs of the MIDT ( Figure 1A), as described in our previous studies [47,48]. Across subject they completed an average of 184 ± 4 (mean ± SD) trials. Each trial starts with a bet (a dollar, a cent, or n money). After a randomized interval of 1-5 s, a target box is presented and disappears after response window. Subjects are requested to make a response as quickly as possible to collect th money (win) before it disappears, following by feedback that indicates the amount of money wo (in red) or lost (in blue). (B) Accuracy rate and (C) RT of trials (mean ± SD).

Imaging Data Preprocessing and Group Analyses
Briefly, brain images were collected using multiband imaging (multiband factor = with a three-Tesla MR scanner (Siemens Trio, Erlangen, Germany). Data were analyze with Statistical Parametric Mapping (SPM8, Wellcome Department of Imagin After a randomized interval of 1-5 s, a target box is presented and disappears after a response window. Subjects are requested to make a response as quickly as possible to collect the money (win) before it disappears, following by feedback that indicates the amount of money won (in red) or lost (in blue). (B) Accuracy rate and (C) RT of trials (mean ± SD).

Imaging Data Preprocessing and Group Analyses
Briefly, brain images were collected using multiband imaging (multiband factor = 3) with a three-Tesla MR scanner (Siemens Trio, Erlangen, Germany). Data were analyzed with Statistical Parametric Mapping (SPM8, Wellcome Department of Imaging Neuroscience, University College London, UK), including realignment, slice timing, co-registration, segmentation, normalization, and smoothing, as in our earlier studies [39,48]. We examined event-related BOLD signals in a single model focusing on the feedback or outcome phase of win or loss processing, as described in our previous study [47]. We performed one-sample t tests of win vs. nil and loss vs. nil. To investigate the neural correlates of AUDIT and BDI-II, we conducted whole-brain linear regressions of these contrasts on AUDIT and Brain Sci. 2022, 12, 1689 4 of 15 BDI-II, separately, with age, sex, and FTND scores as covariates. All models were evaluated with a threshold combining voxel p < 0.001, uncorrected, and cluster p < 0.05 family-wise error (FWE), corrected, following current reporting standards. Voxels with peak activity were indicated with Montreal Neurological Institute (MNI) coordinates. We performed inclusive masking to identify the neural correlates shared between AUDIT and BDI-II for win vs. nil and loss vs. nil, respectively.

Mediation Analyses
We examined how activations of the regions of interest, AUDIT and BDI-II scores were inter-related with mediation analyses [49], as described in our prior work [50]. The mediation test was performed by employing three regression equations [49].
where a, b, c , and c represent path coefficients, and variable M is a mediator of the correlation X → Y. The significant paths a and b, as well as (c-c ), indicate that X → Y is mediated by M. Moreover, if the path c is not significant, then X → Y is completely mediated by M. Table 1 summarizes the demographic and clinical characteristics for men and women separately. The mean FTND score was <1, suggesting a largely non-or light-smoking sample. No sex differences were noted for age, years of education, AUDIT, BDI-II, or FTND score; thus, we combined men and women in data analyses. Figure 1B,C show the accuracy rate and reaction time (RT) of dollar, cent, and nil trials. The accuracy rates were close to 67%, suggesting the success of the staircase procedure. Across subjects, the AUDIT score was significantly and positively correlated with the BDI-II score without (r = 0.555, p < 0.001) or with (r = 0.560, p < 0.001; Figure 2) age, sex, and FTND as covariates. The AUDIT or BDI-II score did not show a significant correlation with either accuracy rate or with RT of dollar, cent, and nil trials (p's ≥ 0.062 without covariates; p's ≥ 0.109 with age, sex, and FTND as covariates), or the differences of dollar/cent vs. nil in either accuracy rate or RT (p's ≥ 0.338 without covariates; p's ≥ 0.308 with age, sex and FTND as covariates).

Brain Activations of Win vs. Nil and Loss vs. Nil
In one-sample t-tests, we evaluated regional activations to win vs. nil and loss vs. across all subjects. The results are shown in Supplementary Figure S1 and summarized Supplementary Table S1. Compared to nil, win trials showed higher activations in bilateral caudate, anterior cingulate cortex, bilateral lingual gyri, and cerebellum; l trials showed higher activations in the bilateral lingual gyri, anterior cingulate cor bilateral insula, bilateral precentral gyri, and cerebellum.

Whole Brain Regressions on AUDIT and BDI-II Scores
We performed whole-brain regressions of win > nil ( Figure 3A,B; Table 2) and los nil ( Figure 4A,B; Table 2) against AUDIT and BDI-II scores, separately, with age, sex, a FTND score as covariates across subjects. We identified the clusters that overlapp between AUDIT and BDI-II regressions ( Figures 3C and 4C). For win > nil, overlapping clusters included bilateral parahippocampal gyrus, superior frontal gyr and posterior cingulate cortex/precuneus, left inferior parietal lobule and infer temporal gyrus, and right middle temporal gyrus, putamen, and insula. For loss > nil, overlapping clusters included bilateral hippocampus/parahippocampal gyrus, super frontal gyrus, mid-cingulate cortex, cerebellum, and right amygdala, middle tempo gyrus, thalamus, and inferior parietal lobule.

Brain Activations of Win vs. Nil and Loss vs. Nil
In one-sample t-tests, we evaluated regional activations to win vs. nil and loss vs. nil across all subjects. The results are shown in Supplementary Figure S1 and summarized in Supplementary Table S1. Compared to nil, win trials showed higher activations in the bilateral caudate, anterior cingulate cortex, bilateral lingual gyri, and cerebellum; loss trials showed higher activations in the bilateral lingual gyri, anterior cingulate cortex, bilateral insula, bilateral precentral gyri, and cerebellum.

Whole Brain Regressions on AUDIT and BDI-II Scores
We performed whole-brain regressions of win > nil ( Figure 3A,B; Table 2) and loss > nil ( Figure 4A,B; Table 2) against AUDIT and BDI-II scores, separately, with age, sex, and FTND score as covariates across subjects. We identified the clusters that overlapped between AUDIT and BDI-II regressions ( Figures 3C and 4C). For win > nil, the overlapping clusters included bilateral parahippocampal gyrus, superior frontal gyrus, and posterior cingulate cortex/precuneus, left inferior parietal lobule and inferior temporal gyrus, and right middle temporal gyrus, putamen, and insula. For loss > nil, the overlapping clusters included bilateral hippocampus/parahippocampal gyrus, superior frontal gyrus, midcingulate cortex, cerebellum, and right amygdala, middle temporal gyrus, thalamus, and inferior parietal lobule. Brain Sci. 2022, 12, x FOR PEER REVIEW 6 of 16

Mediation Models
The AUDIT and BDI-II scores were positively correlated, as shown earlier. As expected, the activation (β) of overlapping clusters was each positively correlated with the AUDIT score (r = 0.66, p < 0.001 for win > nil and r = 0.65, p < 0.001 for loss > nil) and BDI-II score (r = 0.67, p < 0.001 for win > nil and r = 0.64, p < 0.001 for loss > nil). Thus, we conducted mediation analyses to examine the relationships amongst the βs, AUDIT, and BDI-II scores. We tested all six models for each contrast. The results showed that the β of win > nil and of loss > nil each completely mediated the relationship between BDI-II and AUDIT ( Figure 5). None of the other models showed significant mediation at a corrected threshold p < 0.05/6 = 0.0083 (Table 3).

Mediation Models
The AUDIT and BDI-II scores were positively correlated, as shown earlier. As expected, the activation (β) of overlapping clusters was each positively correlated with the AUDIT score (r = 0.66, p < 0.001 for win > nil and r = 0.65, p < 0.001 for loss > nil) and BDI-II score (r = 0.67, p < 0.001 for win > nil and r = 0.64, p < 0.001 for loss > nil). Thus, we conducted mediation analyses to examine the relationships amongst the βs, AUDIT, and BDI-II scores. We tested all six models for each contrast. The results showed that the β of win > nil and of loss > nil each completely mediated the relationship between BDI-II and AUDIT ( Figure 5). None of the other models showed significant mediation at a corrected threshold p < 0.05/6 = 0.0083 (Table 3).

Figure 5. Mediation models: brain activation (β) of (A) win > nil (in green) and (B) loss > nil (in red)
of the shared ROIs completely mediated the correlation between BDI-II (in blue) and AUDIT (in yellow) scores. All six models were assessed for each contrast and evaluated at a corrected threshold p = 0.05/6 = 0.0083. The p values associated with mediation are for the path "a × b" (see Section 2). The statistics of all models are summarized in Table 3. Table 3. Mediation models of AUDIT, BDI-II score, and brain activation (β) each for win > nil and loss > nil.  The statistics of all models are summarized in Table 3. Table 3. Mediation models of AUDIT, BDI-II score, and brain activation (β) each for win > nil and loss > nil.
p Values

Discussion
We identified regional brain responses to monetary win and loss outcomes in correlation with both AUDIT and BDI-II scores in non-dependent drinkers. Individuals with higher AUDIT and BDI-II scores showed greater activation to wins in bilateral frontoparietal cortex, precuneus/posterior cingulate cortex, and right temporal cortex. Individuals with higher AUDIT and BDI scores also showed higher activation to losses in bilateral (but predominantly right-hemispheric) hippocampus/parahippocampal gyrus, cerebellum, right temporal and inferior parietal cortex, and posterior cingulate cortex. These regional activities completely mediated the relationship between depression and alcohol use severity. The findings highlight shared neural correlates of reward and punishment processing between depression and alcohol misuse and may help research of the etiologies of comorbid depression and AUD. We highlight the major findings for discussion.

Depression and Alcohol Misuse Shared Neural Responses to Monetary Win and Loss
A wide array of cortical and subcortical regions was involved during win and loss processing in link with depression and problem drinking. Most notable among these regions are bilateral hippocampi/parahippocampal gyri (HC/PHG), which are shared for both contrasts-win vs. nil and loss vs. nil-although the latter also involved the right amygdala in activities shared between depression and alcohol misuse. The HC/PHG is best known for its function in memory encoding and retrieval, and high-arousing, salient events consistently engage the HC/PHG [51,52]. Relative to nil, both win and loss trials are more salient; thus, the current findings suggest elevated HC/PHG responses to saliency both in association with the severity of depression and alcohol misuse. These findings are broadly consistent with previous reports of HC/PHG structural and functional changes in depressive and anxiety [53][54][55] and alcohol use [56][57][58] disorders. Studies of the etiological mechanisms of depression have emphasized ill-adaptive HC/PHG circuit responses to stress [59]. An imaging literature has associated with HC/PHG circuit dysfunction in emotional and reward processing [60][61][62]. Here, we demonstrated that depression and alcohol misuse both implicate the HC/PHG in heightened responses to wins and lossessalient stimuli irrespective of their valence.
The shared correlates that appeared to be specific to win and loss processing are the posterior cingulate cortex and precuneus (PCC/Pcu) and mid-cingulate cortex (MCC), respectively. Although less of a focus in human studies of reward processing, the PCC is implicated in post-decisional reward signaling in neuronal recordings from behaving monkeys [63]. Notably, the PCC/Pcu did not show significantly higher responses to win vs. nil in the one-sample t test (Supplementary Figure S1), suggesting this default mode network region solely as a correlate of individual variation in depression and alcohol misuse. In contrast, a hub of the limbic motor circuit, the MCC responds to learning of aversive consequences and behavioral avoidance [64,65]. The MCC along with other midline brain regions, including the supplementary motor area, showed significantly higher activation to loss vs. nil in the one-sample t test (Supplementary Figure S1). Thus, individuals with more severe depression and alcohol misuse would engage the MCC greater than the average extent, likely to support emotional motor processes of negative reinforcement. Also notable is the rostral anterior cingulate cortex (rACC), which showed higher responses to both win and loss vs. nil across subjects but only higher responses to win vs. nil in correlation with AUDIT but not BDI-II score. The rACC is part of the saliency and executive control circuit [66,67]; rACC responses to reward may potentially represent a unique marker of alcohol misuse. These findings support studies of neuromodulation of the ACC as a treatment of AUD [68,69]. More research is warranted to evaluate sub-regional cingulate cortical responses to reward and punishment and how these responses may be altered in depressive and alcohol use disorders [70].

Shared Neural Responses Mediated the Link of Depression and Alcohol Misuse
Across individuals, BDI-II and AUDIT scores were positively correlated, supporting the comorbidity of depression and problem drinking even in non-dependent drinkers [7,10,[71][72][73][74]. Importantly, we observed shared brain activities in response to reward and to punishment, and these activities completely mediated the relationships between the severity of depression and alcohol misuse. Specifically, with correction for multiple testing, the findings suggest that depression contributes to alcohol misuse through the shared neural responses to monetary win and loss. As discussed in the Introduction, earlier studies have suggested individual differences in reward sensitivity as a risk factor of both depression and alcohol misuse [75][76][77]. The current findings expand this literature by showing evidence that depression contributes to alcohol drinking through brain activities in response to reward and punishment. Notably, although only one of the mediation models-depression → β of win or loss → problem drinking-showed significant, complete mediation, the model problem drinking → β of win or loss → depression showed an incomplete mediation effect with a p < 0.009. Given the small sample size of the study, it remains to be seen whether the shared correlates may mediate the link of depression and problem alcohol use bidirectionally.

Limitations of the Study and Future Directions
Several limitations need to be noted for the study. Firstly, the study comprised a small sample and, though evaluated at a corrected threshold, the findings and particularly those of regional responses shared between AUDIT and BDI-II regressions, need to be replicated. For the same reason, we did not investigate potential sex differences in the current findings or whether depression and alcohol misuse may be associated with shared reduction in neural responses differentiating reward and punishment [78]. Secondly, reward and punishment come in many forms. Thus, the findings should be considered specific to monetary win and loss. It remains unclear whether or how neural processes of other modalities of reward and punishment (e.g., social interaction and rejection) may be shared in the etiological processes of depression and alcohol misuse. Lastly, although the participants in the current study were largely non-or light smokers (mean FTND score < 1), we cannot entirely rule out the effects of smoking on the current findings. Studies of a larger sample and perhaps of individuals with varying levels of nicotine dependence would be needed to address the effects.
Although the study focused on reward processing, depression and alcohol misuse involve and may share other etiological processes, including fronto-limbic dysfunction in impaired inhibitory control and emotion processing and learning [79][80][81][82][83] and altered mitochondrial bioenergetics [84][85][86]. More studies are needed to evaluate the respective and potentially interactive roles of these neural and biological mechanisms of depression and alcohol misuse. Further, to the extent that these findings are considered specifically for non-dependent drinkers, future studies may incorporate assessment of physical, social, and cultural factors that may influence alcohol intake and the emotional context under which individuals engage in use and misuse of alcohol and "alcohol-like" drinks [87,88].

Conclusions
To conclude, we showed a significant correlation between the severity of depression and alcohol misuse in non-dependent drinkers. Neural responses to monetary wins and losses both mediated the relationship between depression and problem drinking. Suggesting shared etiological processes, these findings not only enhance our understanding of the neural mechanisms associated with both psychiatric conditions but also provide evidence for the importance of concurrent treatment of depression and alcohol misuse in clinical populations.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/brainsci12121689/s1, Figure S1 Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data and codes will be shared on request to the corresponding author.