LPFC Organization in Emotion-Duration Difference Estimation

Overview

To support optimal behavior in daily life, goals and responses following emotional events should ideally incorporate not only the valence and intensity of prior emotional episodes but also their temporal features, such as the relative duration of positive vs. negative attributes. However, how specific brain regions contribute to the integration of temporal and emotional information and promote goal-directed response remains unknown. The goal of this study is to examine how specific brain regions track both emotional and temporal information of dynamic emotional events to inform other related brain regions to guide goal-oriented and context-appropriate actions. The investigators will scan healthy human participants using functional MRI (fMRI) while they view emotional image sequences and track the associated emotional and temporal (duration) information, and act accordingly. The investigators will employ multivariate patterns analysis and pattern similarity analysis to identify brain regions that represent (can decode) emotion, time, and their combined signals, as well as brain regions that represent the associated action goal. In addition, to infer the causal contributions of these brain regions in forming task-relevant representations (emotion, time, and action goal), the same participants will be recruited to receive transcranial magnetic stimulation (TMS) in these regions.

Overview. n=50 participants from the UC Santa Barbara and the larger Santa Barbara community will be recruited and invited to participate in a multi-session fMRI/TMS+fMRI study. Eligibility criteria have been described and include fMRI and TMS safety criteria. Behavioral Task. To test the hypothesis that the lateral frontal pole (FPl) represents emotion and temporal signals and integrates them to inform contextual action goal representations in the mid-lateral prefrontal cortex (mid-LPFC), the investigators will use a well-validated task that manipulates time and emotional valence factors to inform context-dependent action goals, the Emotional Sequences Task. In each trial, participants view 12-s sequences of 4 novel negative and positive images (events). Half the trials feature longer-lasting negative (vs. positive) picture presentations, and vice versa, yielding greater temporal evidence for either positive or negative emotional valence. The amount of temporal evidence in favor of one valence in a 12-s sequence (∆ Temporal evidence: 1200ms vs 1800ms) is varied orthogonally with respect to the predominant emotional valence by varying individual picture presentation times (2000ms-4000ms; jittered). At the end of each sequence, participants indicate whether the total duration of positive or negative events was longer with a button press (Left vs. Right), following an action-mapping rule (Contextual Goal Cue). Procedure. fMRI acquisition. Using a 64-channel coil in the 3T Prisma Siemens MRI scanner located at UCSB's Brain Imaging Center (BIC), the investigators will collect whole-brain EPIs (multiband factor=3; 2.5 mm3 isotropic voxels; TR=1.5 s; TE=30 ms; FA=65°) and T1-weighted images for spatial normalization and TMS neuronavigation (0.94 mm3; TR=2.5 s; TE=2.19 ms; FA=7°). fMRI data processing and modeling. Following our prior work (see(2)), fMRI data processing, in FSL and Python, will include motion and slice time correction, 3mm FWHM smoothing, and alignment to T1-space maintaining native functional resolution. Trial-wise BOLD activation parameters for emotional sequence and action preparation epochs will be obtained using a canonical HRF in a Least-Squares All (LS-A)(3) GLM, and regularized using multivariate noise normalization(4) to remove nuisance inter-voxel correlations due to physiological and/or instrument noise. Regions of Interest (ROIs). PFC-LPFC: FPl and mid-LPFC (BA46 \& 9-46); mPFC: BA25 and BA32-Oxford PFC consensus atlas(5-7) and surface-based segmentation in Freesurfer(8) aligned to individual T1 space. Amygdala ROIs-CITI atlas(9) and ANTs registration to T1 space. fMRI analysis: Overview. Aims 1-1b: the investigators will model the following factors: predominant emotional valence (positive vs. negative; hereafter, emotion), # temporal evidence (1200ms vs. 1800ms; hereafter, time), and action goals (rule-based left vs. right; hereafter, action). MVPA. To replicate our prior work(2), which showed linear decoding of emotional valence in FPl, and action goals in FPl and mid-LPFC, the investigators will use a multivariate logistic classifier run on z-scored data for each subject, ROI, and task epoch (emotional sequences and action preparation) in Nilearn. Classifier performance will be evaluated using the area under the curve (AUC; chance performance = 0.5) and leave-one-run-out cross validation (yielding run-wise classifier AUCs). Mixed Models: Overview. Here and throughout, the investigators use mixed models (lme4(10)) entering subject and run as random factors. Classifier Mixed Models. Classifier AUC is tested against chance using mixed models combining run-wise classifier AUCs across subjects. Representational Similarity Analysis (RSA). Full-factorial RSA tests whether the inter-trial similarity structure of multivoxel neural activity patterns (i.e., the neural similarity matrix) is explained by experimental factors-here, emotional valence, time, action goals, and, critically, their interaction(2). The investigators will obtain neural similarity matrices for each participant, ROI, and task epoch by computing Pearson correlations between pairs of trial-wise multivoxel patterns in a between-runs correlation approach, which minimizes inflated correlations due to data dependencies and autocorrelation. RSA Mixed Models. Next, the investigators will fit a multiple regression mixed model for each ROI using condition-specific template matrices (each template matrix is included in the subject error term). The investigators will fit emotion, time, and emotion\*time regressors (emotional-sequences epoch). This model tests whether time and emotion significantly explain the similarity structure of FPl and mid-LPFC neural activity patterns, and, critically, whether they interact (i.e., in conjunctive representations(11,12)). To test whether mid-LPFC and FPl represent action goals, the investigators will fit an action goal regressor (action preparation epoch); a secondary model will include emotion and time as interactive regressors (emotion\*action, time\*action, emotion\*time\*action) to examine whether action goal representations are unaltered by valence in mid-LPFC (vs. FPl) (as the investigators previously found(2)) or whether these factors are integrated in mid-LPFC in this task. The behavioral relevance of time, emotion, and goal signals in LPFC. The investigators will test the predicted behavioral import of integrated time-emotion signals in FPl by regressing the fit of emotion\*time regressors in FPl (RSA betas) on accuracy in the Emotional Sequences Task. FPl-mid-LPFC interactions: Functional connectivity and cross-regional representational evidence. The investigators will test whether FPl's integrated time-emotion signals may inform mid-LPFC function by regressing (a) FPl's emotion\*time fits (RSA betas) and (b) FPl-mid-LPFC's functional connectivity (PPI betas) on mid-LPFC's action-goal signals (AUC). Statistical Rigor. Here and throughout, all p values are FDR-adjusted to correct for multiple comparisons. Regional and representational specificity in LPFC is formally tested by entering region (FPl vs. mid-LPFC) as an interactive regressor in above-described mixed models. The investigators will bring participants from Experiment 1.1 (Aim 1) for 3 TMS+fMRI sessions targeting FPl, mid-LPFC, and a non-PFC Control (S1) (order counterbalanced across subjects). Each TMS administration will be followed by fMRI scanning of the Emotional Sequences Task. Experiment 1.2 tests a causal role of FPl function in (a) informing action-goal representations in mid-LPFC and (b) task performance (requiring accurate tracking of temporally-extended emotional information); compared to Experiment 1.3, which targets action-goal signals in mid-LPFC, and Experiment 1.4, which targets a non-PFC control (S1). Procedure. Information-guided TMS. Following others'(13) and our(1) recent work, individualized LPFC TMS sites will target the peak location of task-relevant classifier evidence- emotional valence in FPl, and action goals in mid-LPFC-as revealed by a spherical searchlight (7.5mm radius) run on fMRI data obtained at baseline (Experiment 1.1; Aim 1) constrained by relevant anatomical ROIs (see ROIs, Aim 1). S1 targets are defined based on anatomy(14-17). TMS sites will be a-priori restricted to the Left hemisphere given more consistent engagement of Left vs. Right LPFC during cognitive control(18-20) and emotion regulation(15,21). TMS protocol. TMS will be delivered using a Magstim Horizon Lite magnetic stimulator and a figure-of-eight coil. Precise TMS targeting is achieved using T1-weighted scans and computerized stereotaxic system (Brainsight). As in our prior work, the investigators will use an offline cTBS protocol (50Hz trains of 3 pulses every 200ms; 40 s; Huang et al. 2005), which reduces cortical activity for up to 60 min after stimulation. Questionnaires. The investigators will assess mood pre and post TMS using the PANAS(22) and STAI(23). fMRI analysis. The investigators will conduct MVPA and RSA (detailed in Aim 1). Statistical Rigor. See Aim 1; here, TMS site specificity is tested by entering TMS site (FPl vs. mid-LPFC vs. Control/S1) as an interactive factor in above-described mixed models (classifier AUC and RSA); all p-values FDR corrected for multiple comparisons.