Dynamic cognitive characteristics of emotional con ict in subthreshold depressed patients: An Event-Related Potential Study


 Subthreshold depression (SubD) has a considerable impact on an individual’s subjective well-being and psychosocial functioning, and is a risk factor for Major depression disorder (MDD). The inability to effectively control and resolve emotional conflict is a typical symptom of certain mood disorders, and the aim of this study was to confirm impairments in cognitive processing mechanisms for emotional conflict processing in SubD patients with event-related potential (ERP) recording. The study of the mechanisms of emotional conflict in subthreshold depression may provide an ideal model for understanding the neurophysiological mechanisms and developing preventive strategies in patients with MDD. Methods:The Healthy control (HC) and SubD groups were recruited, with 32 subjects in each group completing the word-face Stroop paradigm, during which ERP amplitudes and latencies were recorded. Results:Compared to HC group, the SubD group had lower accuracy and longer response times in both the "consistent stimulus" and "inconsistent stimulus" conditions. Regardless of the stimulus condition, the SubD group had a greater N2 amplitude in the prefrontal mid-lobe region. In the SubD group, the N450 amplitude was also found to be greater in the prefrontal middle region for the "incongruent stimulus minus congruent stimulus" and the conflict SP amplitude was smaller in the parieto-occipital region for the "incongruent stimulus minus congruent stimulus". Conclusions:The findings suggest that, supported by behavioural and brain evidence, people with SubD have dynamic cognitive deficits in emotional conflict processing, specifically greater sensitivity to early processing of emotional stimuli and sharper detection of emotional conflict, but more delayed adaptation and response options following emotional conflict resolution.


Introduction
Major depression disorder (MDD) has a high rate of disability and death and has become a major cause of worldwide burden of disease (Ferrari et al., 2013). It is critical to determining the factors that contribute to the increased risk of MDD. Patients with subthreshold depression (SubD) have the similar symptoms as those with depression, but to a lesser degree and do not meet the criteria for a diagnosis of depression (Dai Q et al., 2012;Hayasaka Y et al., 2015). Because of its high prevalence, insidious symptoms (Lee, Y. et al., 2019), and signi cant negative impact on patients' quality of life (Tuithof, M. et al., 2018), SubD is increasingly becoming a greater burden on health services than MDD. Adolescents are a vulnerable population for SubD (Bertha and Balazs, 2013). Research shows that SubD affected 32% of Chinese university students (Jiang et al., 2019) and 23%-39% of European college students (Mikolajczyk et al., 2008). SubD is a risk factor for depression (Bertha and Bala zs,2013). SubD are more likely to develop depression than normal people (Cuijpers and Smit, 2004). A longitudinal study showed the risk of developing SubD into MDD was as high as 10-25% within 1-3 years compared to the normal population (Cuijpers and Smit, 2004). Therefore, SubD research may provide a superb model for Comprehension the pathophysiology of MDD and further for developing preventive strategies for MDD.
MDD results from abnormalities in emotional processing mechanisms within the brain and manifests as de cits in emotional neurocognitive processes (Disner SG et al., 2011). Speci cally, MDD selectively perceives and evaluates the emotional information around it (Disner SG et al., 2011;Everaert J et al., 2012), triggering abnormally high levels of negative emotions and abnormally low levels of positive emotions (Jones, E. B. and Sharpe, L., 2017). Besides this, There is research evidence that MDD suffers from con ict monitoring dysregulation, which is re ected in competitive tasks regarding response options. (Ottowitz WE et al., 2002;Epp AM et al., 2012). This suggests that dysregulated con ict monitoring may be a marker of executive dysfunction in people with MDD. However, con ict monitoring in life is not only cognitive, many emotionally con icting stimuli in the environment can also interfere with our ongoing behaviors (LeDoux JE, 2000;Tipples J and Sharma D, 2000).
Emotional con ict develops from cognitive con ict and is the interference of unrelated stimuli with the current cognitive task (Etkin A et al., 2006 ). Research suggests that the inability to resolve and adapt to emotional con icts is typical of people with certain mood disorders, such as depression and anxiety disorders (Williams JM, 1996;Xue S, 2017). The study found that compared to healthy controls, people with depression had longer reaction times and lower accuracy in naming the colours of sad words in an emotional stroop paradigm (Ottowitz WE et al., 2002). Therefore, defective emotional con ict processes are an underlying de cit in Depressed people (Anderson ND et al., 2007), and controlling emotional con ict is important for both normal functional performance and treatment of mood disorders. The research on emotional con ict in depressed patients is comprehensive and in-depth, both at the neurological and behavioural mechanism levels (Epp AM et al., 2012), but relatively little research has been done on SubD patients. Research on emotional con ict in patients with SubD is important not only for uncovering the mechanisms of emotional con ict, but also for diagnosis and treatment.
Most previous studies examining emotional con ict have used the emotional stroop paradigm ,but there is no real emotional con ict formed between the emotional meaning of words and the names of colors in this paradigm (Algom D et al., 2004). The word-face Stroop paradigm, which is transformed from the emotional Stroop paradigm, is one of the main paradigms for studying emotional con ict and can directly assess emotional con ict. It operates by marking emotional words in emotionally colored faces and asking subjects to judge the emotional value of the face or word (Stenberg G et al.,1998;Kavcic V and Clarke J M, 2000). Previous studies have shown that the event-related potential (ERP) technique is used widely. It is a technique with high temporal resolution that compensates for the shortcomings of other methods in cognitive neuroscience and allows us to better understand cognitive processes.
A series of ERP studies using the word-facet Stroop paradigm found that word-facet incongruent stimuli triggered a more negative N2 component around 200-300ms post-presentation, and that N2 amplitude in the frontal central region may be related to emotion control and con ict detection (Thomas S J et al., 2007;van Hooff J C et al., 2008). In addition, word-face incongruent stimuli triggered a more negative N450 component in the central frontal region of the midline 300-500ms after presentation, which is thought to show con ict detection in the Stroop task (Coderre, Conklin and van Heuven, 2011). The wave amplitude was greatest in the left of midline region for N450 in the early phase (350-450ms) and in the central midline region for N450 in the late phase (450-500ms), because early N450 may come from a left frontal location, unlike early N450, late N450 originate from the central frontal and ACC. this suggests that early and late components of N450 may be involved in different stages of cognitive control, respectively (Taake I et al., 2009;Wei D et al., 2010). It has also been found that the con ict slow potential (SP) is a con ict-sensitive positive component over parietal areas during the 700-800ms time window, and we usually associate the SP with con ict resolution (N. DePisapia and T.S. Braver, 2006), response selection (A.T. Chen et al., 2011) and monitoring (M.J. Larson et al., 2009).
Synthesizing existing research, we decided to use erp technology and adopt a word-faced stroop paradigm to study mechanisms of dynamic cognitive processes in patients with SubD during an emotional con ict task. In terms of behavioural outcomes, we assumed the SubD group would have longer reaction time (RT) and lower accuracy (ACC) in the task compared to the HC group. At the neurological level, we hypothesize that SubD patients exhibit abnormal neurophysiological indicators, such as an overall increase in N2, N450 and SP amplitude levels in the wordface dysregulation condition, where SubD patients exhibit interference effects. This suggests that impaired neural processes underlie the cognitive control de cits associated with SubD.

Materials And Methods
Participants 64 undergraduate students (42 women, 22 men) were recruited from the Shandong University of traditional Chinese medicine Jinan China which through advertisements and grouped according to their scores on the Center for Epidemiological Survey. Depression Scale (CES-D) (Radloff LS, 1977)and the Hamilton Depression Scale-17 (Hamilton M, 1967). Both scales had Chinese adaptations and had been widely used. The SubD group were screened using a cutoff score of≥16 on the CES-D according to previously reported literature (Buntrock C et al., 2017), which is the usual way of indicating "having depressive symptoms" in CES-D. The Hamilton Depression Scale-17 was then used to con rm SD and exclude clinical depression using a score range of between 7 and 17 indicating a mild depression (0~6 no depression, 7~17 mild depression 18~24 moderate depression above 25 severe depression) (Ballesteros J et al., 2007). The Hamilton Depression Scale-17 was performed by at least two assessors, at least one of whom was quali ed as a national psychological consultant. All assessors had received training in the assessment's administration. The inclusion criteria for the HC group were CES-D score of 16. Exclusion Criteria for both SubD and HC participants were (1) did not ful ll the SCID diagnostic criteria for MDD; (2) had no current bipolar disorder, panic disorder or schizophrenia; (3) had no concurrent psychotherapy and psychotropic medication; and (4)not pregnant and currently not in their menstrual period.
All participants were right-handed, and at the time of the experiment, the SubD group was not receiving medication or was entering the experiment 1 month after weaning from medication. Prior to participating in the experiment, subjects were carefully informed of the purpose of the experiment and the procedure and signed an informed consent form. The study was conducted in accordance with the Declaration of Helsinki as revised in 1989. The subjects were compensated after completing all study procedures.

Experimental material
The stimulus materials comprised both happy-face and sad-face photographs of people. All photographs were selected from the Native Chinese Affective Picture System (Lu B et al., 2005), which has been deemed to be appropriate for use with the Chinese population. We selected 20 happy-face and 20 sad-face photographs according to a 1:1 sex ratio and used red font for the words " (' gaoxing' means happy)" or " ('beishang' means sad)" over the central portion of the faces in the photographs. The text appears on the nose area of the face and the font size is approximately 1cm x 1cm, an example of which is shown in Figure 1. In the task, there are two conditions (congruent, incongruent), congruent in the sense that the facial expression in the picture agrees with the meaning of the word, and incongruent in the sense that the facial expression does not agree with the meaning of the word The image was eventually rendered on a 17 inch CRT computer screen, measuring 5cm by 7cm.

Experimental Paradigm
The word-face Stroop paradigm was used in this study. For the speci c experimental design, refer to the study by Song Xue et al (Xue, S et al., 2017). The entire experiment comprised one practical trial and four blocks, with each block consisting of 80 trials. After completing a block, the participants were provided 1-2 minutes of rest. When the experiment started, one photograph containing an emotional face and an emotional word was shown in the middle of the screen for 1500ms. The stimulus time interval was between 3000ms and 5000ms (average: 4000ms), and the central screen displayed a "+" between the stimuli.
For the congruent and incongruent tasks in this study, the experimenters asked the participants to determine whether the face was sad or happy and to press a button in response as soon as possible. The participants were instructed to press the "S" button with their left-hand nger when the face appeared sad and the "K" button with their right-hand nger when the face appeared happy. The order of the stimuli was pseudorandom, and all experimental stimuli, as well as the sex of the faces, were balanced. The picture presentation was ended once the participant pressed the correct button or after 1500ms had elapsed. A blank screen was then presented for 800 to 1200ms. In total 160 trials were recorded for each type of stimuli. The whole experiment took about 30 minutes.
The experiment was conducted in a quiet environment, using two computers, one for presenting the stimuli and one for recording the EEG data. During the experiment, subjects were asked to minimise body movements, especially head and facial movements. The subjects were asked to look at the screen with their eyes, their faces 75 cm away from the screen and their horizontal and vertical viewing angles less than 68 degrees.
In order to ensure that the subjects were su ciently familiar with the whole procedure, a 20-trial pretest was conducted before the formal start of the experiment, and the pre-test procedure was the same as in the formal experiment.

EEG Recording
In this experiment, a 64-channel electrode cap is used to continuously record EEG. The electrode cap is distributed according to the international 10/20 system, and the sampling rate is 512Hz. After the original signal is ampli ed, a bandpass lter (0.01-100Hz) is applied. EEG data were collected using a Biosemi (Biosemi, UvA, NL) ActiveTwo Data Conversion Box. Impedances were kept below 5 kΩ throughout the recording. The preprocessing and analysis of the EEG data was performed in Matlab (version R2014a; MathWorks, Inc, MA, USA) using the EEGlab toolbox (Delorme and Makeig, 2004), the ERPlab toolbox (Lopez-Calderon and Luck, 2014).
The electrodes were referenced to the CMS (Figure 2.) during acquisition and the REST was used as the o ine reference, with the study shows that it is more appropriate to use REST reference scheme settings in facial recognition experiments . Electrodes were placed above and below the left eye and at the outside of the left and the right canthi to measure the electrooculogram. Ocular artefacts were corrected using the independent component analysis (ICA) procedure in the EEGlab toolbox (Delorme and Makeig,2004). The o ine high-pass lter is set to 0.1 and the low-pass lter is set to 35 Hz, with a band-pass lter to improve the signal-to-noise ratio. The data 200 ms before the start of the stimulus is used for the baseline correction of the ERP. The continuous EEG is divided into the period starting -200 ms before the start of the stimulus and ending at 1000 ms after the start of the stimulus. Tests exceeding ±100μV are excluded from further analysis. Epochs containing artefacts were rejected prior to averaging. The average number of accepted trials was 145 (SD=6) in the word-face emotion consistency condition, 151 (SD=4) in the word-face emotion inconsistency condition.
To quantify the ERP data, we calculated mean amplitudes for each condition in 150~250ms, 350~450ms and 700~800ms time windows. We chose the above time window based on previous studies and visual inspection of ERP waveforms. We included the 150~250ms time window to capture the N2 effect, the 350~450ms to capture the N450 effect, and the 700~800ms time window to capture the SP effect. Based on previous research and observation of the difference wave topography. In this study, N2 and N450 in the central prefrontal region and SP in the posterior central region of the parieto-occipital lobe were selected for analysis. Three electrode sites, FZ, FPZ and FCZ, were selected for analysis with average wave amplitudes of N2 and N450 in the central prefrontal region, and three electrode sites, PZ, POZ and OZ, were selected for analysis with average wave amplitudes of SP in the posterior central parieto-occipital region.Please see Figure 3 for details.

Statistical analysis
Descriptive data were presented as mean±SD unless otherwise speci ed. T-test and X 2 test were used to compare the demographic data and depression scales (CES-D and Hamilton Depression Scale-17) between SubD participants and HCs.
The E-prime procedure was used to record subject reaction time (RT)of correctly and the accuracy of responses (ACC) spss 22 to analyse these behavioural data. A two factor repeated measure ANOVA with RT and ACC as dependent variables, respectively, was used to examine the group (SubD vs HC) as the between-subject factor, and word-face emotional coherence (congruent vs incongruent) as the within-subject factor. The signi cance level was 0.05.
A three-factor repeated measure ANOVA was used to examine the latencies and amplitudes of N2, N450 and SP, with the group (SubD vs HC) as the between-subject factor, and word-face consistency (congruent vs incongruent) ×electrode point (N2 (FZ, FPZ, FCZ); N450 FZ, FPZ, FCZ ;SP PZ, POZ, OZ) as the within-subject factor. The normality of data was checked before applying parametric tests. The Greenhouse-Geisser correction was used to correct repeated measure ANOVA that violated the sphericity assumption, and the Bonferroni correction was applied to all post hoc pairwise comparisons. The signi cance level was 0.05.

Demographic Characteristics
Demographic data for subjects in the HC group and SubD group are shown in Table 1. The mean age was 19.813±0.859 years in the SubD group and 19.563±0.759 years in the HC group. The two groups were no statistically signi cant differences on the age(t=-1.233; P=0.222), gender(X 2 =0.549; P=0.443) and education(t=0.151; P=0.880) dimensions, showing that the two groups were well matched. In contrast, as an important difference in our setting, CES-D scores are signi cantly different between HC group.and SubD group (t=22.504; P 0.001).

Analysis of Behavioral Data
The mean and SD of RT and ACC for each condition are shown in Table 2. First, RT was used as a dependent variable to analyze the repeated measures of ANOVA for the group (SubD vs HC) × word-face emotion consistency (congruent vs incongruent). The results showed that the main effect of group was signi cant (F (1,31) =22.361, P<0.001), and the main effect of word-face emotion consistency was signi cant (F (1,31) =55.346, P<0.001), and the interaction between them was signi cant (F (1,31) =25.432, P<0.001) further simple analysis of the effect shows that the SubD group's RT under the condition of word-face emotion inconsistency was signi cantly greater than the word-face emotion consistency (F (1,31) =8.242, P<0.001), and the HC group's RT under the condition of emotional inconsistency is also signi cantly greater than the word-face emotional agreement F (1,31) =3.312, P<0.001).
Then, ACC was used as a dependent variable to analyze the repeated measures of ANOVA for the group (SubD vs HC × word-face emotion consistency (congruent vs incongruent). The results showed that the main effect of group was signi cant (F (1,31) = 5.690, p<0.001), and the main effect of word-face emotion consistency was signi cant (F (1,31) = 96.916, p < 0.001), other effects were not signi cant. After pair-wise comparison, it was found that the ACC of SubD group was lower than that of HC group, and under consistent was higher than that under inconsistent.

N2
The latencies and amplitudes of N2 were statistically analyzed. Please see Table 3 and Figure 4

N450
The amplitudes and latencies of N450 were statistically analyzed. Please see Table 4 and Figure 4 for details. The ANOVA on N450 amplitudes showed a main effect of electrode site (FZ=-1.511μV; FCZ=0.350μV; FPZ=0.219μV; F(2,62)=21.783, P<0.001 with the largest amplitudes in the FZ and the smallest amplitudes in the FCZ. There was a main effect of group (HC=0.694μV SubD=-1.322μV; F(1,31)=8.136, P=0.008<0.05) with smaller amplitudes in the HC group than in the SubD group. There was a main effect of word-face emotion consistency (congruent=-0.112μV; incongruent=-0.517μV; F (1,31) =20.615, P<0.001), and it was found that there was a trend interaction between group and word-face emotion consistency (F (1,31) =5.051, P=0.032<0.05), further analysis of the simple effect showed that the HC group had a smaller differential wave of N450 induced by discordant and consistent conditions than the SubD group (differential wave of HC group=-0.210μV; differential wave of SubD group=-0.600μV; P=0.001 0.001).

SP
The amplitudes and latencies of SP were statistically analyzed. Please see Table 5  incongruent=1.107μV; F (1,31) =36.627, P<0.001), and it was found that there was a trend interaction between group and word-face consistency (F (1,31) =12.283, P<0.001), further analysis of the simple effect showed that the SubD group had a smaller differential wave of SP induced by discordant and consistent conditions than the HC group (Differential wave of HC group=1.347μV; Differential wave of SubD group=0.543μV; P 0.001).
The ANOVA on SP latencies showed all the effective effects are not signi cant.

Discussion
Emotional con ict is a cognitive process in which we need to select our target information and suppress irrelevant information when processing emotional information, and the resolution of emotional con ict has great signi cance in real life. This study examined the behavioral and electrophysiological correlates of emotional con ict de cits in SubD patients.
The study found that subjects showed a signi cant behavioural Stroop interference effect. Both the SubD and HC groups had longer RTs in the incongruent condition than in the congruent condition when completing the emotional con ict task. This is consistent with the ndings of Botvinic (Botvinick MM et al., 2004), who identi ed stronger activation in the prefrontal as well as the cingulate cortex in the incongruent condition, showing a signi cant decrease in behavioural indicators during response. Previous research (Etkin et al., 2006) has suggested that this is due to a con ict between emotional and cognitive resources. When word-face emotional valence is in con ict, word and face competition for cognitive resources, such as attention, and the resources taken up by processes such as recognition and processing of non-target stimuli can interfere with processes such as recognition and processing of target stimuli, and in turn make them less e cient. This study also found that the SubD group had signi cantly longer RT and lower ACC rates in both conditions compared to the HC group, suggesting that SubD may have de cits in the cognitive processes of emotional con ict.
Regarding the ERP data, the SubD group had greater N2 amplitude in the frontocentral region and greater N450 difference wave amplitude in the frontocentral region for "incongruent stimuli minus congruent stimuli" relative to the HC group. The SubD patients and HC group showed enhanced con ict SP for "incongruent stimuli" than for "congruent stimuli," and the HC group showed greater SP difference wave amplitude for "incongruent stimuli minus congruent stimuli" than the SubD patients. This, combined with the longer RT and lower ACC behavioural data during the task, suggests that SubD patients have neurological impairments in monitoring and responding to emotional con ict.
For N2, the subject type main effect was signi cant, showing a larger wave amplitude in the SubD group than in the HC group. N2 re ects the state of attention at the beginning of the entire cognitive processing, and the results suggest that SubD patients are more sensitive to the early processing of emotional stimuli than normal individuals. The present study did not nd differences in the con ict condition, but previous studies found that the N2 amplitude of incongruent stimuli was greater than the N2 amplitude of congruent stimuli in the interference task (Gehring WJ et al., 1992;Kopp B et al., 1996;Holmes AJ et al., 2008), possibly because previous studies have mostly used the classical Stroop or go-nogo paradigms, whereas the present study used the word-face Stroop paradigm. Previous studies (Zhang JP et al., 2021) have shown that people with SubD pay different attention to emotional information than normal people and engage with emotional information to a greater extent, so the N2 amplitude was more negative in the SubD group, re ecting the fact that people with SubD require more cognitive resources to process emotional information.
For N450, the main effect of con ict type was signi cant, and waves in the incongruent condition had a greater turn in the frontal mid-lobe region than waves in the congruent condition. N450 is a valid indicator of the Stroop interference effect (Shen YM et al., 2013 ), and the ndings suggest that there was a signi cant Stroop interference effect in this study. The main effect of subject type was signi cant, with the SubD group eliciting more negative N450 compared to the HC group, and an interaction effect between subject type and con ict type was found, with further simple effects analysis showing that the incongruent condition minus the more congruent condition elicited greater N450 in the SubD group than in the HC group.The N450 component can be used as an electrophysiological indicator of con ict processing e ciency (Strommer-Davidovich N et al., 2018), which is positively correlated with the amount of psychological resources an individual devotes to con ict events (West R et al., 2005), and this study shows that people with SubD have low con ict processing e ciency and high interference sensitivity. It has also been suggested that N450 may be related to con ict monitoring (Liotti M et al., 2000;West R et al., 2003) and con ict resolution. For example, Qiu et al. (Qiu NJ et al., 2006) in a colour-word Stroop study using Chinese characters as materials found N450 responded to con ict monitoring and con ict resolution processes, so in the present study we suggest N450 re ects participants' monitoring and resolution of con ict between facial expressions and the meaning of emotional words associated with the facial word Stroop task. The greater wave of difference in N450 in the SubD group implies that SubD patients are more sensitive to the detection of emotional con ict.
For SP, a main effect of con ict type was signi cant, with incongruent conditions triggering greater SP than congruent conditions. Subject type main effects were evident, with the SubD group triggering smaller SP amplitudes than the HC group, which differed from the hypothesis, and an interaction effect between subject type and con ict type was found, with the SubD group inducing smaller waves of difference in SP in incongruent conditions minus congruent conditions than the HC group. Many studies (Whitney C et al., 2009;Clayson PE et al., 2011;Forster SE et al., 2011) have linked the slow wave of con ict SP to post-response monitoring and con ict adaptation. For example, Chen et al. (Chen A et al., 2011) suggest that con ict SP is related to post-response monitoring because this positivity extends beyond the mean response time. Larson et al. (M.J. Larson et al., 2009) suggest that post-response SP re ects the con ict adaptation process. Therefore, we suggest SP responds to subjects' monitoring of responses and con ict adaptation in the completion of an emotional con ict task. The results of the present study imply that adaptation and response monitoring are more delayed in people with SubD following emotional con ict resolution.
Research has shown that the inability to resolve and adapt to emotional con ict is typical of patients with certain mood disorders. The present study provides new behavioural and neurophysiological evidence for impaired emotional con ict control in patients with SubD. The ndings suggest that, compared to the HC group, SubD patients have abnormal emotional con ict inhibition using the Word-Face Stroop task, as evidenced by greater sensitivity to early processing of emotional stimuli, overcommitment of cognitive resources, and sharper detection of emotional con ict, but delayed adaptation and response selection following emotional con ict. On the one hand, these ndings help us to enhance our understanding of SubD inhibition defects, on the other hand, they also provide evidence on the cognitive neuropsychiatric model of SubD.

Statement
All subjects complete an informed consent form and agree to publication in an online open access publication.