Deviations From Normative Functioning Underlying Emotional Episodic Memory Revealed Cross-Scale Neurodiverse Alterations Linked to Affective Symptoms in Distinct Psychiatric Disorders

Abstract

Background: Affective symptoms are a prevalent psychopathological feature in various psychiatric disorders. However, the underlying neurobiological mechanisms are complex and not yet fully understood.

Methods: We used normative modeling to establish a reference for functional activation of functional magnetic resonance imaging based on an emotional episodic memory task, which is frequently used to study affective symptoms in psychiatric disorders. This normative reference was derived from a large dataset of healthy individuals (n = 409) and used to evaluate individualized functional alterations by calculating deviations from this reference in a clinical dataset, which included 164 healthy control participants and patients with major depressive disorder (MDD) (n = 56), bipolar disorder (BD) (n = 31), and schizophrenia (SZ) (n = 73). The functional deviations were mapped to emotional networks (ENs) with specific emotional functions and used to predict affective symptoms in different mental disorders. The microscale cellular signatures underlying macroscale variations were identified using imaging transcriptomic analysis and associated with affective symptoms.

Results: We observed distinct patterns of cross-scale neural alterations linked to affective symptoms in 3 psychiatric disorders. Macroscale neural dysfunctions in distinct disorders were embedded into non-overlapping ENs and significantly associated with affective symptoms. Oligodendrocytes may mediate the network-specific impairments and microglia for MDD, astrocytes for BD, and excitatory neurons for SZ as replicable cell-type correlates of affective symptoms.

Conclusions: These findings revealed cross-scale neural alterations underlying affective symptoms in psychiatric disorders, providing a basis for understanding their neuropathological patterns and guiding individualized treatment.

Keywords: Affective symptoms; Cellular decoding; Neuroimaging; Normative modeling; Psychiatric disorders; Task fMRI.

Figure 1

The workflow of the current study design. (A) Data on functional activation for each participant under task–fMRI of emotional episodic memory were extracted from 2 research datasets (the healthy dataset and the clinical dataset). (B) The normative model was first established based on functional activation of each region as response variables and task performances as predicted covariates, drawing from the HC_train group of the healthy dataset. The HC_test group from the clinical dataset was further put into the normative model to replicate validation, and individualized alterations were evaluated as deviations relative to the normative range in any given brain region for each patient in the clinical dataset. (C) Overview of analysis pipeline comprising 3 steps. Model generalizability was assessed based on the similarity between the HC_train and HC_test groups; the stability of measurements derived from the normative reference was evaluated through the replicate validation step using different brain parcellation atlases and modeling methods and validated whether these measurements could be influenced by the medication effect (step 1). The macroscale alterations were quantified as the differences between the proportion of individuals showing an extreme deviation of each case and HC participants at the regional level and further embedded into emotional network for specific functional annotations of emotional profiling. Individual deviations with significant differences were used to predict affective symptoms of each psychiatric disorder (step 2). Transcriptomic analysis and cell-type enrichment were further applied to examine the microscale alterations related to macroscale variability and linked to affective symptoms in distinct disorders (step 3). BD, bipolar disorder; fMRI, functional magnetic resonance imaging; HC, healthy control; MDD, major depressive disorder; ROI, region of interest; SZ, schizophrenia.

Figure 2

The emotional episodic memory task and generalizability evaluation of the normative model. (A) The emotional episodic memory task consisted of encoding and retrieval phases, and each phase included neutral and aversive scenes. The participants were instructed to determine whether each image represented an indoor or outdoor scene in the encoding phase (left) and to respond “new” or “old” in the retrieval phase (right). (B) The distribution of mean regional deviations in the HC_train and HC_test groups. There was no significant difference between these 2 distributions. (C) The spatial distribution of the regional deviation map in the HC_train group exhibited significant spatial correlations with the group-averaged activation and regional deviation maps in the HC_test group. HC, healthy control.

Figure 3

Regional differences and clinical symptom associations in patients with major depressive disorder (MDD), bipolar disorder (BD), and schizophrenia (SZ). The infranormal and supranormal regions in patients with MDD (A), BD (D), and SZ (G) were identified using 10,000 times group-based permutation test, and statistical significance was adjusted by false discovery rate (FDR). Regional differences were further enriched into 4 large-scale emotional networks (ENs). The infranormal and supranormal alterations at the network level in the patients with MDD (B), BD (E), and SZ (H) are shown. The statistical significance was determined by using 10,000 times group-based permutation test and adjusted by FDR. The scatter plot showed the significant associations between individual functional deviations and clinical affective symptoms in patients with MDD (C), BD (F), and SZ (I). HAMD, Hamilton Depression Rating Scale; HC, healthy control; PANSS, Positive and Negative Syndrome Scale; RMSE, root-mean-square error.

Figure 4

Imaging transcriptomic analysis and cell-type enrichment. (A) Schematic diagram of imaging transcriptomic analysis at the group and individual level, respectively. Infra- and supranormal Δ percentage maps for each diagnostic group were spatially correlated with the 15,633 gene expressions of 6 postmortem brain tissue samples from the Allen Human Brain Atlas (AHBA) using Spearman correlation analysis. Individual deviation maps from each patient with a psychiatric disorder were also spatially correlated with AHBA gene expressions. Based on 2 independent cell-type gene markers, the analysis of cell-type enrichment was conducted using the fast gene set enrichment analysis (FGSEA) method in the sorted gene rank to separately calculate group cell-type enrichment and individual-level cell-type enrichment scores. (B) Group-level cell-type enrichment. Cell types enriched by infra- and supranormal deviations are shown across 2 cell-type gene markers. The horizontal bar plots showed the statistical significance of enrichment that was quantified by sign(normalized enrichment score [NES]) × −log₁₀(p-value) given by the FGSEA method. The replicable significant cell types across 2 datasets are labeled. The network group effects of individual deviations corresponding to significant cell-type enrichment are also marked on the side. (C) Symptom associations of individual cell-type enrichment. The associations between individual-level NES of specific cell type and clinical affective symptoms were examined using the Pearson correlation analysis in distinct psychiatric disorders. The cell types that were significantly enriched at the group level also exhibited significant associations with clinical affective symptoms in each psychiatric disorder. The bar plots show the statistical significance of correlations, and replicable significant cell types across 2 datasets are labeled. BD, bipolar disorder; EN, emotional network; MDD, major depressive disorder; OPC, oligodendrocyte precursor; SZ, schizophrenia.