Decreased activity in the reward network of chronic insomnia patients

In modern society, many people have insomnia. Chronic insomnia has been noted as a risk factor for depression. However, there are few functional imaging studies of the brain on affective functions in chronic insomnia. This study aimed to investigate brain activities induced by emotional stimuli in chronic insomnia patients. Fifteen patients with primary insomnia and 30 age and gender matched healthy controls participated in this study. Both groups were presented images of fearful, happy, and neutral expressions consciously and non-consciously while undergoing MRI to compare the activity in regions of the brain responsible for emotions. Conscious presentation of the Happy-Neutral contrast showed significantly lower activation in the right orbitofrontal cortex of patients compared to healthy controls. The Happy-Neutral contrast presented in a non-conscious manner resulted in significantly lower activation of the ventral striatum, right insula, putamen, orbitofrontal cortex and ventral tegmental area in patients compared to healthy controls. Our findings revealed that responsiveness to positive emotional stimuli were decreased in insomniac patients. Specifically, brain networks associated with rewards and processing positive emotions showed decreased responsiveness to happy emotions especially for non-conscious image. The magnitude of activity in these areas also correlated with severity of insomnia, even after controlling for depression scale scores. These findings suggest that insomnia induces an affective functional disorder through an underlying mechanism of decreased sensitivity in the regions of the brain responsible for emotions and rewards to positive emotional stimuli.


Scientific Reports
| (2021) 11:3600 | https://doi.org/10.1038/s41598-020-79989-2 www.nature.com/scientificreports/ for certain suspected diseases: the PSQI 38 and AIS 39 for insomnia, Epworth Sleepiness Scale 42 for hypersomnia, STOP-Bang questionnaire 43 for sleep apnea, the International Restless Legs Syndrome Study Group rating scale for the severity of restless legs syndrome 44 and parasomnia, the Munich Parasomnia Screening 45 for sleep behavior disorder, and the Rapid Eye Movement Sleep Behavior Disorder Screening Questionnaire 46 for rapid eye movement sleep behavior disorder. A sleep-wake cycle survey by actigraphy and sleep apnea screening and pulse measurement using pulse oximetry were conducted for the first night at home for one week. Participants revisited the laboratory one week after the tests at home to fill the same questionnaires as the PT group, undergo the MMSE, and perform the emotional-face viewing task. The tests were conducted between 10:00 and 15:00.
The exclusion criteria for the healthy participants were as follows: those with any of the following sleep disorders-insomnia, hypersomnia, sleep apnea, parasomnia, rapid eye movement sleep behavior disorder, or circadian rhythm sleep disorder; any serious physical conditions, use of medication or other substances that may affect the results of the study (hypnotics, drugs that induce sleepiness such as anti-histamines and steroids); psychiatric diseases; implanted metal devices such as pacemakers; ophthalmic conditions including abnormal color perception; individuals doing shift work; individuals who had traveled abroad to a region with a 6-h or greater time difference in the past 3 months; individuals with 200 mg/day or more caffeine intake, heavy smokers (those who would find it stressful to abstain from smoking for 5 days); and patients with suspected cognitive decline (an MMSE score of ≤ 25). Emotional-face viewing task. The emotional-face viewing task developed by one of the authors of this study 47 was conducted while the participants underwent MRI as shown in Fig. 1. The participants were presented images of emotional facial expressions under the following conditions: (1) a conscious condition wherein the emotional facial expression image was presented for long enough for the participant to visually recognize it consciously or (2) a non-conscious condition, wherein the emotional facial expression image was presented for too short a duration for the participant to consciously recognize it, following which it was masked with a neutral expression. The 48 images used as stimuli in this study were selected from two standardized facial stimuli sets (Ekman and Friesen 48   . Emotional-face viewing task. Facial pictures depicting fearful, happy, or neutral expressions were used as the stimuli and were presented either non-consciously or consciously. In a non-conscious trial, an emotional image (either fearful, happy, or neutral) was implicitly presented for 26 ms, followed by an explicit presentation of a neutral "masking" face of the same individual as the preceding implicit emotional face (when the implicit face was neutral, the following explicit mask was of a different person of the same sex) for 174 ms. Participants were required to press a button in response to each "target" stimulus to keep themselves awake during the scanning. www.nature.com/scientificreports/ atr-p.com/produ cts/face-db.html), and included fearful, happy, and neutral faces from four Japanese and four Western individuals of both genders. For both the conscious and non-conscious conditions, a point image was presented for 1000 ms to guide the participants' gaze. Following this, for the conscious condition, one of the facial expressions (fearful, happy, or neutral) was presented for 200 ms. For the non-conscious condition, one of the facial expressions was presented for 26 ms, after which the image was backward-masked by a neutral face from the same individual. When a neutral image was presented first, an image of the neutral expression from a different individual of the same sex was presented for backward masking. Images were presented in blocks of eight trials. The target stimulus was presented randomly during each block, and participants were asked to press a button to react. During the rest block, the gazing point image was presented for 15 s. Each of the two sessions consisted of the following sequence repeated six times with different kinds of facial expressions: non-conscious condition block → conscious condition block → rest block. The second session was conducted after a 1-2-min break.
After the completion of each session, the participants were asked to report their subjective sleepiness during the task on the following scale: 0 = Not sleepy at all, 1 = Slightly sleepy, 2 = Sleepy, 3 = Very sleepy, and 4 = I fell asleep for a little bit. The mean scores of the subjective sleepiness for the two sessions were calculated for analysis.
After completing the two sessions, the participants were asked to report how they saw the masked face using the following scale: 0 = It was completely unrecognizable, 2 = I saw something, but I was not sure what kind of expression it was, 3 = I was able to recognize the expression in some of the pictures, and 4 = I recognized the expression in almost all the pictures.
There were no participants who answered 3 or 4 with regards to how they saw the masked face.
Functional MRI data acquisition. A  ac.uk/spm/softw are/spm8/) was used for the brain functional imaging data analyses. Functional images were motion and slice timing-corrected, co-registered to a magnetization prepared-rapid gradient echo structural image, spatially normalized to the Montreal Neurological Institute template, and smoothed with an 8 mm fullwidth half maximum Gaussian kernel. The time-series data including the 3D blood-oxygen-level-dependent signals of the participants were analyzed with the general linear model (first-level fixed model effect). The timeseries of each session was convolved with the canonical hemodynamic response implemented function in SPM8, and the functions of the hypothesized hemodynamics for the various conditions of the stimulus presentation were created. The time-series of the blood flow models for the responses to the six emotional and conscious stimulus categories (non-conscious/conscious happy, non-conscious/conscious fear, and non-conscious/conscious neutral) and the time-series data of the variables related to the six body movements were built into the design matrix as a regressor. Actual blood-oxygen-level-dependent signals were analyzed voxel by voxel with a general linear model, and the beta values corresponding to each regressor were calculated. For inter-participant testing, the beta values between the conditions of interest for each voxel were analyzed using a paired t-test. The contrast (activity from viewing the neutral expression image subtracted from the activity from viewing the emotional expression image) was then created. A total of four contrasts for the fear-neutral and happy-neutral expressions presented consciously and non-consciously were generated.
Areas known to have important associations with positive or negative emotions, i.e., the amygdala, insula, anterior cingulate cortex, and areas involved in responding to pleasurable stimuli or motivation for reward [50][51][52][53][54][55][56][57] , were selected as regions of interest. Masks for the right and left amygdala, right and left insula, anterior cingulate cortex, corpus striatum (putamen + caudate nucleus), and right and left orbitofrontal areas were created using the Anatomical Automatic Labeling from the SPM WFU PickAtlas toolbox (Wake Forest University; http://fmri. wfubm c.edu/downl oads/WFU_PickA tlas_User_Manua l.pdf). A mask for the ventral tegmental area, for which a corresponding item does not exist in the Anatomical Automatic Labeling, was created with a 5-mm radius sphere centered on the peak site, as described in a previous study 58 . A covariance analysis was performed with sex, age, sleepiness, and recognition of non-conscious images as covariates for voxels within the masks to compare the various conditions between the HC and PT groups. A small volume correction for the multiple comparisons was performed at the peak level within the masks. The risk level for the multiple comparison correction was set at a family-wise error of 5% at peak level.
MRI partial correlation analysis: severity of insomnia and depression scale scores. The sites with significant differences in the non-conscious happy-neutral contrasts were extracted using MarsBaR, and their correlations with the AIS, PSQI, and SDS scores were studied. There were two clusters in the right insula; thus, the cluster with higher statistical values was used for analysis. Furthermore, SDS scores were significantly higher in the PT group. A partial correlation analysis was performed to show that the high SDS scores in the PT group were attributed to insomnia and not to a spurious correlation with high depression scores to assess whether the correlations with the PSQI and AIS were still significant after controlling for the severity of depression as measured using the SDS. Multiple comparison correction was performed by adjusting the false discovery rate: FDR (Benjamini and Hochberg 59 method).

Results
Basic parameters of the participant groups. The data of each participant group are shown in Table 1.
There were no significant differences in age, sex ratio, subjective sleepiness, recognition of the masked image, or button-pressing reaction time between the two groups. The PSQI, AIS, and SDS scores were significantly higher in the PT group than in the HC group.
Comparison of the fMRI data. To investigate the functional changes in the brains of the insomnia patients, we compared the activity in the regions of their brain responsible for emotions to that in healthy controls. The sites with significant differences and their statistical values are shown in Table 2. There were no significant differences between the groups in brain activity in response to the conscious presentation of fear in any of the observed regions.
Compared to the healthy controls, insomnia patients had significantly lower activity levels in the middle and posterior parts of the left insula in response to the non-conscious presentation of fear ( Fig. 2: p < 0.05, small volume correction). There were no areas of the brain with more activity in the insomnia patients than in healthy controls.  Table 2. Brain regions with statistically significant differences in activation between insomniacs and healthy controls (p < 0.05, small volume correction). BA Brodmann area, MNI Montreal Neurological Institute template. Cluster k*p < 0.001 threshold. † p < 0.05 threshold (p-value for the ventral tegmental area cluster was greater than 0.001, but significant at peak level small volume correction); HC = healthy control group; PT = insomnia patient group. www.nature.com/scientificreports/  www.nature.com/scientificreports/ The activity in response to the conscious presentation of happiness was significantly decreased in the right orbitofrontal cortex of insomnia patients compared to that of healthy controls ( Fig. 3; p < 0.05, small volume correction). There were no significant differences in the brain activity in response to the facial expressions presented under other conditions. There were no sites showing a greater activity in the insomnia patients than in healthy controls.
In response to the non-conscious presentation of happiness, the activity in the ventral striatum, putamen, ventral tegmental area, bilateral orbitofrontal cortices, and the right anterior insula was significantly lower in insomnia patients than in healthy controls (Fig. 4: p < 0.05, small volume correction). There were no significant differences in the brain activity in these regions in response to the presentation of facial expressions in other conditions. There were no sites with greater activity in the insomnia patients than in healthy controls.
Correlation analysis. The activity in the areas associated with reward had a significant negative correlation with the PSQI and AIS scores. The ventral tegmental area also showed similar correlations, though the correlation was not significant. SDS scores had a significant negative correlation with the activity in the right orbitofrontal cortex and ventral striatum, but not with that in the right insula (Table 3). All significant results survived after FDR correction. Brain activity in response to the non-conscious presentation of happiness. The map shows a significantly greater activation in response to non-conscious happy face stimuli in the healthy control than the insomnia patient group. Significant differences were observed in the ventral striatum, putamen, orbitofrontal cortex (OFC), right insula, and ventral tegmental area (VTA). Significant clusters are rendered on a T1 anatomical referential image displayed in neurological convention, with the left side corresponding to the left hemisphere. The cluster shown on the VTA were made using a threshold with a lenient alpha level (p < 0.05, k > 5) for visualization purposes. Table 3. Correlation matrix analyses between the brain activity and scales of insomnia and depression. L left side, R right side, OFC orbitofrontal cortex, VTA ventral tegmental area, vSTR ventral striatum, PSQI The Pittsburgh Sleep Quality Index, AIS Athens Insomnia Scale, SDS Self-Rating Depression Scale. † p < 0.1, *p < 0.05, **p < 0.01. Degrees of freedom (df) = 45. www.nature.com/scientificreports/ A partial correlation analysis with SDS scores as a control variable showed that the correlations between the brain activity and severity of insomnia were significant for all areas of the brain studied (Table 4). All significant results survived after FDR correction.

Discussion
The present study revealed that the activity in the brain's reward system (corpus striatum, ventral tegmental area, orbitofrontal cortex, ventral tegmental area, and right insula) in response to facial expressions of happiness were decreased in insomnia patients relative to that in controls. This is the first study to demonstrate that the changes in cerebral functions in response to positive emotions are altered in insomnia patients. Furthermore, we found that these patterns were observed after controlling for the symptoms of depression and that the severity of insomnia had a strong negative correlation with the activity in these brain regions.
The corpus striatum comprises of the putamen, caudate nucleus, nucleus accumbens, and other structures, and is well known to be associated with motivation and mood modulation 60,61 . The activity in the corpus striatum has been strongly associated with stimuli involving positive valence or reward, and this activity correlates strongly with the subjective assessment of valence [51][52][53] . The corpus striatum and other areas, such as the orbitofrontal area and insula, are also involved in goal-directed behaviors motivated by food or monetary rewards [54][55][56][57] . A metaanalysis of reward-related tasks 62 has shown that the corpus striatum is associated with both the prediction and consumption of the reward. The corpus striatum receives dopaminergic projections from the ventral tegmental area of the midbrain and forms the neural networks responsible for the motivation, reinforcement, and reward processes with the surrounding areas, such as the insula, anterior cingulate cortex, and amygdala 63 . Activity in this dopaminergic system has been documented to play an important role in modulating the brain activity in response to positive rewards 60,61 .
The activation of the amygdala has been reported in several studies of non-conscious happy facial expression presentation tasks 64,65 . Many previous studies have reported that the presentation of non-conscious emotional facial expressions can elicit behavioral and physiological responses 66 , such as judgement modulation 67 , a modification of consumption behavior 68 , and non-consciously synchronized facial expressions 69 . The difference in the activity level between the amygdala and anterior cingulate cortex in response to the non-conscious presentation of happy faces versus sad faces has been reported to decrease with antidepressant therapy 70 . Though only a small number of studies have investigated the responses in the corpus striatum and insula, Chen et al. showed that the ventral striatal activity in response to the non-conscious (masked) presentation of an image of a close friend's face smiling predicted subsequent developments in the friendship, and that it seemed to reflect the activity in the reward system 71 . The present study is the first to reveal that activity in the amygdala and ventral striatum in response to positive expression (happy) stimuli presented non-consciously is decreased in insomnia patients.
Only a few fMRI studies have investigated the reward processes in insomnia patients. One study assessed the brain activity in adolescent girls as they performed a monetarily rewarding task and detected no changes in the activity in the reward system in the brain, but found that the activity in the dorsomedial prefrontal cortex could be a mediator of the relationship between insomnia and depressed mood 72 . Wang et al. reported an altered corpus striatum resting-stage functional connectivity between the default mode network and the sensorimotor network in primary insomnia patients 73 . Patients with a history of bipolar disorder have decreased activity in the ventral striatum and orbitofrontal cortex in response to the conscious presentation of happy expressions compared to healthy individuals 74 . Decreased amygdala activity has been observed in patients with severe depression in response to a happy expression stimulus, but this returned to normal levels during remission 75 . Furthermore, activity in the amygdala and nucleus accumbens in response to happy expressions is lower in children and young adults at a high risk of depression 76 . Anhedonia in patients with depression correlates negatively with the amygdala and ventral striatum activity associated with the presentation of happy expressions 77 . These examples show that the activity of the neural networks related to emotions and rewards is decreased in individuals with, or at risk of, depression. The similarities between the reactions of the insomnia patients in these studies and those in the present study may be related to the lack of motivation and high depressive mood during the day in insomnia patients, and their higher risk of developing depression. Moreover, for the non-conscious presentation of happiness, the severity of insomnia predicted the activity of these regions, even after controlling for depression scores. This finding shows that the difference between insomnia patients and healthy individuals observed was not a spurious correlation attributed to the highly depressive state of insomnia patients, but supports the hypothesis that insomnia is directly linked to abnormal affective functions. Patients with chronic insomnia have been shown to exhibit an excessive craving for sleep and abnormal attention to sleep-related stimuli 78,79 . Dai et al. 80 reported that the hyperactivity of the value-based attentional networks was associated with an excessive attention to sleep-related stimuli. Furthermore, an association between sleep craving and impaired function of the exploratory system, including the reward system, has been reported, and the association between these www.nature.com/scientificreports/ modulations of brain function and hyperarousal in chronic insomnia patients has been discussed. Moreover, the association between the exploratory system and the value-based attentional networks was shown to be reduced in insomnia patients than in healthy controls. Our findings of reduced activity in the reward system are also consistent with these findings; this effect may be related to the pathogenesis of hyperarousal in insomnia, mediated by the impairment of the proper allocation of reward-based attention in these patients, causing them to pay excess attention to sleep-related stimuli. We expected to find overactivity of the emotion-related areas in insomnia patients due to hyperarousal; however, contrary to our expectations, there was no increased activation of the emotion-related areas in the PT group compared with the HC group. Conversely, we found reduced activity in the reward network, suggesting that the hyperarousal in insomnia is not associated with an overactive state of the entire brain. In this study, the middle and posterior parts of the left insula were significantly less activated by the non-conscious presentation of fearful faces in insomnia patients than in the control group. A meta-analysis on the functions of the insula 81 has suggested that there is a strong relationship between affective tasks and the right insula, but associations with the middle part of the left insula have not been confirmed. However, significant activity in this region has been observed in studies on somatosensory and motor functions. The decreased activity in the middle left insula in response to negative stimuli is likely explained by the altered physiological functioning 8-10 associated with the hyperarousal in insomnia patients.
A significant difference between the two groups in the brain activity in response to the presentation of fear was only observed when the presentation was non-conscious. Conscious presentation of happy expressions generated a between-group difference in the orbitofrontal cortex only, while a difference was observed in a broader area of the reward system when the presentation was non-conscious. There are two hypothesized reasons for these differences in brain activity. First, the non-conscious presentation may have generated simple responses from the neighboring systems that were unaltered by the visuo-cognitive or high-order functions of the prefrontal lobe. Indeed, a previous study reported that the magnitude of the amygdalar response to non-consciously presented emotional expressions was a strong predictor of personality traits and subjective sleepiness 47,82 . Etkin et al. investigated the strength of the correlation between the amygdalar activity and intensity of anxiety when participants consciously or non-consciously viewed an expression of fear and found significant correlations between the intensity of anxiety and amygdalar activity in response to a non-conscious presentation. They attributed the high correlation between the non-conscious processes and anxious traits to the notion that, unlike conscious processes, the non-conscious presentation is not affected by an inhibitive response to the presentation of an image 82 . The second predicted reason is the ceiling effect. While the difference between the two groups in their conscious responses to images were negligible, the non-conscious presentation of images may have provided the optimal stimulus intensity to observe a difference between the two groups.
This study has several limitations. First, data on the subjective assessments of the facial expression stimuli were not collected. Thus, it is unknown whether the stimuli induced pleasant emotions in the participants. Because the assessment of emotional stimuli has been reported to lower the activity in the brain regions responsible for emotions 83,84 , we did not ask participants to make subjective assessments while performing the task. However, the activity of the reward system decreased in response to non-conscious presentation; therefore, we assume it to be a system that is not directly related to subjective pleasure. This indicates that subjective assessments would not be relevant to the results of this study.
The second limitation is related to the sampling method. Although the patients who participated in this study encompassed a wide age range, the mean age was approximately 60 years; thus, our findings may not be directly comparable to those of previous studies that have used younger participants. A sample with a broader age range should be selected for further investigations.
Third, in this study, the PT group had 3-day washout period for hypnotics which was managed by the attending physician. It is possible that the patients in this group experienced symptoms of rebound insomnia. These symptoms might enhance the differences in brain activity between the two groups.

Conclusion
The present study revealed a decreased responsiveness to positive emotional stimuli in regions associated with emotions and rewards in insomnia patients. This is the first report to show changes in cerebral functions related to positive emotions in insomnia patients, compared with healthy controls. A reduced responsiveness to positive stimuli is also observed in patients with depression or bipolar disorder and may be associated with the highly increased risk of depression in insomnia patients. Future studies should compare insomniacs to depression patients, as well as study whether similar patterns in responsiveness are also observed in people with non-clinical insomnia. Insomnia patients often show increased activity of the sympathetic nervous system [11][12][13] and higher levels of physiological markers of stress such as cortisol [14][15][16][17][18] , thus the relationships between these markers and brain activity should be investigated as well.