Mike Yassa
A. Research Question, Goal, or Specific Aims
We have developed and published a pipeline for determining structural and connectivity data for the paraventricular nucleus (PVT) using 3T MRI data (Kark et al. 2021). The PVT has been established as an important circuit in regulation of conflict behavior as well as is critical for long-term fear memory formation. These processes are integral to PTSD-related mechanisms (i.e. trauma memory and recovery, avoidance behavior). We propose to utilize this new pipeline to test 2 hypotheses in the ENIGMA/PCG-PTSD consortium dataset (>possible 3000 PTSD cases and controls).
Hypothesis 1: Structural indices of PVT will be abnormal in PTSD cases vs. controls.
Hypothesis 2: Functional connectivity of PVT will be abnormal in PTSD cases vs. controls. Additional analyses would be to examine association of childhood trauma with PVT structural indices and functional connectivity. We propose to conduct analyses both using group comparison (cases vs. controls) and PVT measures as a function of symptom severity in PTSD symptom clusters, particularly re-experiencing and avoidance clusters. Other variables of Interest, childhood and lifetime trauma burden/load, age, sex, history of substance use, depression symptoms.
B. Analytic Plan
Our processing pipeline is published in Kark et al. 2021. It builds on the minimal processing that is used for Human Connectome Project (HCP) data which includes gradient-nonlinearity-induced distortion correction, rigid body head motion correction, EPI image distortion correction, co-registration between the fMRI and structural data, normalization to MNI space, high-pass filtering (1/2,000 Hz), and brain masking (Glasser et al., 2013) and independent components analysis (ICA)-based artifact removal of noise components from preprocessed fMRI data (ICA-FIX; Griffanti et al., 2014; Salimi-Khorshidi et al., 2014).
Processed images are then input into the CONN Toolbox Version 19c (Whitfield-Gabrieli and Nieto-Castanon, 20122; RRID:SCR_009550). In CONN, the data are segmented (gray matter, white matter, and CSF) and the functional images were spatially smoothed using a 5 mm full width at half maximum (FWHM) smoothing kernel. Artifact identification is performed using Artifact Detection Tools (ART) implemented in CONN. We enforce conservative motion censoring thresholds, scrubbing frames exceeding > 0.5 mm frame-wise motion or Global Signal z > 3. Further denoising is carried out using the CONN Toolbox’s aCompCor method (white matter and CSF noise, frame-wise motion regression, artifact scrubbing, and linear detrending). The data are then band-pass filtered to isolate resting-state frequencies (0.01 Hz < f < 0.10 Hz) using a fast Fourier transform (FFT). Seed regions for the PVT are based on our publicly available masks in Kark et al. 2021 and posted in an Open Science Framework (OSF) repository (https://osf.io/re3v6/).
Multivariate seed-based connectivity (mSBC) is calculated as the semipartial correlation coefficients between the BOLD timeseries of the PVT and all other individual voxel timeseries in the brain after controlling for the BOLD timeseries of the other control seeds. These subject-level connectivity maps are then entered into separate group-level analysis in standardized MNI space. Group functional connectivity statistics are calculated as voxel-level cluster-inferences using TFCE (Smith and Nichols, 2009) implemented in the CONN Toolbox. This will generate a set of subject-level statistical values that can be used as dependent variables in the analyses of interest.
The main independent variable will be based on the diagnosis (PTSD or Control) or symptom severity (PTSD, depression). For group analysis PTSD case description will be based on the Nievergelt et al. 2019 PGC-PTSD Freeze 2 definitions. We will also use PTSD and depression symptoms as continuous measures. The dependent variables will be PVT regional connectivity with specific regions, including the amygdala, the nucleus accumbens, the hippocampus, and the prefrontal cortex. Covariates will include gender, age and site/scanner. Other variables of interest include childhood trauma, trauma burden, and substance use.
C. Analytic Personnel
Danny Stout and Victoria Risbrough will conduct QC and symptom phenotype preparation. Imaging analysis will be conducted by Bianca Leonard, and group analyses will be coordinated between Bianca and Danny. Vickie Risbrough and Mike Yassa will supervise overall analytic strategy and interpretation of results.