Common neural and transcriptional correlates of inhibitory control underlie emotion regulation and memory control

Abstract Inhibitory control is crucial for regulating emotions and may also enable memory control. However, evidence for their shared neurobiological correlates is limited. Here, we report meta-analyses of neuroimaging studies on emotion regulation, or memory control and link neural commonalities to transcriptional commonalities using the Allen Human Brain Atlas (AHBA). Based on 95 functional magnetic resonance imaging studies, we reveal a role of the right inferior parietal lobule embedded in a frontal–parietal–insular network during emotion regulation and memory control, which is similarly recruited during response inhibition. These co-activation patterns also overlap with the networks associated with ‘inhibition’, ‘cognitive control’ and ‘working memory’ when consulting the Neurosynth. Using the AHBA, we demonstrate that emotion regulation- and memory control-related brain activity patterns are associated with transcriptional profiles of a specific set of ‘inhibition-related’ genes. Gene ontology enrichment analysis of these ‘inhibition-related’ genes reveal associations with the neuronal transmission and risk for major psychiatric disorders as well as seizures and alcoholic dependence. In summary, this study identified a neural network and a set of genes associated with inhibitory control across emotion regulation and memory control. These findings facilitate our understanding of the neurobiological correlates of inhibitory control and may contribute to the development of brain stimulation and pharmacological interventions.


Comparison between BrainMap and Neurosynth-based co-activation analysis 29
Co-activation analysis can be performed based on large-scale databases of fMRI studies such as BrainMap 30 or Neurosynth. During the initial data analysis, we took advantages of the relative strengths and 31 weaknesses of the two methods. We augmented the coordinate-based meta-analysis on the Neurosynth 32 with meta-analytic connectivity modelling (MACM) based on the BrainMap to explore co-activation 33 patterns. The idea behind Neurosynth is similar to MACM (which is described in the main text). The 34 Neurosynth algorithm searches for brain regions which co-activate with input coordinates within the same 35 functional contrast, and summarize the results as a co-activation map. Neurosynth has several advantages 36 over BrainMap: (1) Neurosynth is based on automated text mining, therefore includes a higher number of 37 studies and at the same time decreases the potential selection bias of the users.
(2) ROI-based MACM is 38 largely dependent on the size (number of voxels) and the shape of the input ROIs. It could be a potential 39 problem because we included far more studies in the ALE analysis of Stop-signal (SS) paradigm and 40 Go these differences, both methods yielded highly similar co-activation patterns for each ROIs ( Figure S1 and 46 Figure S2), suggesting that the effect of methodology and database on the neural network analysis in our 47 study is negligible. We nevertheless chose presented the results of MACM in the main text. 48 49

Data sharing of non-imaging data via the Open Science Framework (OSF) 50
We used the OSF database (https://osf.io/6wz2j/) as a venue to share non-imaging data generated within 51 this study. 52

Studies and coordinates used in the meta-analyses 53
For each task (e.g. ER: emotion regulation; TNT: think/no-think; GN: go/no-go; SS: stop-signal), an excel 54 file with all coordinates used in the meta-analyses is uploaded to the folder (ALE_coordinates_data) 55 within the OSF. 56

Code 57
Custom python scripts used in this study can be found. 58

Details of MACM results 59
Within the MACM analyses, we estimated in total co-activation patterns of all ROIs from 4 different 60 tasks. Because of the limitation of the number of figures presented in the supplemental materials, we 61 cannot provide all the detailed results for these analyses here. Instead, we used a python package 62 (atlasreader: https://github.com/miykael/atlasreader) to generate coordinate tables and region labels from 63 all co-activation images. All results can be found in an OSF folder (folder name: MACM_results_details, 64 filename: MACM_results_altasreader.zip) 65

Lists of genes 66
Complete gene lists, together with the statistical results, is uploaded in one folder (complete_gene_list) 67 within the OSF database. All files started with "gene_decoding" are the genes whose expression patterns 68 correlated with the brain activity pattern elicited by each task of interest. All files started with 69 "overlapped_gene_list" are genes whose expression patterns correlated with two (ER and TNT) or more 70 (ER, TNT, GN, and SS) task-related activity patterns simultaneously. Numbers (e.g. 500, 1000) were 71 thresholds used in the analysis to select the genes with most similar spatial patterns. All files started with 72 "difference" are genes whose spatial patterns correlated with inhibition tasks, but not DMN. 73

Tables generated by gene ontology enrichment analyses 74
Tables for enrichment analyses of biological functions (GOEA.zip) or diseases items (disease.zip). Results 75 from different thresholds were presented. 76 77

Data sharing and 3D visualization of statistical maps via the Neurovault 78
We uploaded all the statistical maps generated within this study to our Neurovault database 79 (https://neurovault.org/collections/4845/) for data sharing purpose and 3D, interactive visualization. There 80 are in total of 46 maps within the images collection. 81 • Images names starting with the task names (e.g. ER) and contrast names (e.g. Regulation vs 82 Baseline) are maps resulting from the ALE meta-analysis. 83 • Images names ending with threshold methods are corrected ALE images (e.g. ALE C05 1K stands 84 for p<.05, uncorrected p<.001, threshold permutations=1000). 85 • Images names starting with "MACM" are co-activation maps from the MACM analyses. 86 • No images from the coordinate-based co-activation analysis using Neurosynth were uploaded.