Depression in Premanifest HD: Aberrant Effective Connectivity of Striatum and Default Mode Network
Poster No:
261
Submission Type:
Abstract Submission
Authors:
Tamrin Barta1, Leonardo Novelli1, Nellie Georgiou-Karistianis1, Julie Stout1, Samantha Loi2, Yifat Glikmann-Johnston1, Adeel Razi1
Institutions:
1Monash University, Melbourne, Australia, 2University of Melbourne, Parkville, Australia
First Author:
Co-Author(s):
Introduction:
Depression is one of the most common and impactful features early in Huntington's Disease (HD), in the premanifest period (pre-HD), prior to clinical diagnosis (Epping & Paulsen 2011). Depression is increasingly being conceptualised as a circuitopathy (Hannan 2018) and two large-scale networks surmised to contribute to the expression of depressive symptoms in pre-HD are the striatum and the default mode network (DMN; McColgan et al. 2017; Garcia-Gorro et al. 2019). Existing neuroimaging studies are limited and relied on functional connectivity: an inherently undirected measure of connectivity (Friston et al. 2014). Dynamic causal modelling (DCM) allows testing of neurobiologically plausible models of connectivity changes in pre-specified networks (Friston et al. 2014; Razi et al. 2015). We investigated DMN and striatal effective connectivity and depression in pre-HD, using these model-based methods.
Methods:
We analysed 3T resting state fMRI data from 93 pre-HD participants (51.6% females; mean age = 42.7; Klöppel et al., 2015). Behavioural measures included history of depression, Beck Depression Inventory, 2nd Edition (BDI-II) and Hospital Anxiety and Depression Scale, depression subscale (HADS-D). An optimal cut-off score recommended for use in HD categorised clinically significant depressive symptoms (De Souza, Jones, and Rickards 2010).
Regions of interest (ROIs) included medial prefrontal cortex (MPFC [3,54,-2]), posterior cingulate (PCC [0,-52,26]), hippocampus (HPC left [-29,-18,-16], right [29,-18,-16]), caudate (CAU left [-10,14,0], right [10,14,0]), and putamen (PU left [-28,2,0], right [-28,2,0]). Each ROI time series was calculated as the first principal component of the voxels' activity within an 8 mm sphere for MPFC and PCC and a 6 mm sphere for all other regions, and was further constrained within masks. Preprocessing pipeline included slice-timing correction, realignment, spatial normalisation to MNI space, and spatial smoothing by a 6 mm full-width half-maximum Gaussian kernel.
Spectral DCM (Friston et al. 2014; Razi et al. 2015) was used to estimate subject level connectivity and parametric empirical bayes (Friston et al. 2016) was employed to estimate group level effective connectivity changes between participants with a history of depression and those without. We focused on connections that had a Bayesian posterior probability ≥ 0.99. Leave-one-out cross-validation was performed for connections that reached this criterion.
Regions of interest (ROIs) included medial prefrontal cortex (MPFC [3,54,-2]), posterior cingulate (PCC [0,-52,26]), hippocampus (HPC left [-29,-18,-16], right [29,-18,-16]), caudate (CAU left [-10,14,0], right [10,14,0]), and putamen (PU left [-28,2,0], right [-28,2,0]). Each ROI time series was calculated as the first principal component of the voxels' activity within an 8 mm sphere for MPFC and PCC and a 6 mm sphere for all other regions, and was further constrained within masks. Preprocessing pipeline included slice-timing correction, realignment, spatial normalisation to MNI space, and spatial smoothing by a 6 mm full-width half-maximum Gaussian kernel.
Spectral DCM (Friston et al. 2014; Razi et al. 2015) was used to estimate subject level connectivity and parametric empirical bayes (Friston et al. 2016) was employed to estimate group level effective connectivity changes between participants with a history of depression and those without. We focused on connections that had a Bayesian posterior probability ≥ 0.99. Leave-one-out cross-validation was performed for connections that reached this criterion.
Results:
The model estimation was excellent, with an average percentage variance-explained of 89.70% (SD: 4.22; range: 74.15-94.87). For pre-HD with a history of depression, we found excitatory projections from MPFC to right HPC and left PU, in line with expectations (Figure 1). The PCC had aberrant excitatory and inhibitory influence on the striatum and the hippocampus for pre-HD with a history of depression, compared to those without. Striatal connectivity patterns were notable in the more affected left cerebral hemisphere. Contrary to expectations, no aberrant connections were found from MPFC to CAU or PCC. The present study demonstrates that aberrant connectivity patterns for pre-HD with a history of depression is associated with coupling differences in depressive symptoms (Figure 2). Leave-one-out cross-validation comprised left PU, CAU and PCC self-connections, chosen as they appeared most consistently across models. Correct classification reached significance for both HADS-D, corr(91) = 0.19: p = .037, and BDI-II cut-off scores, corr(91) = 0.29, p = .002.
·The weighted colours of the edges represent the estimated effect size and arrows show the directed influence of one region on another, including self-connections.
·A & C represent participants with a history of depression and BDI-II and HADS-D cut-offs respectively. B & D represent participants without depression and BDI-II and HADS-D cut-offs respectively.
Conclusions:
The present study suggests network dysconnection as a neural basis for depression in pre-HD. Aberrant effective connections were associated with trait level depression, which was differentially associated with coupling changes in state depressive symptoms. This work adds to our understanding of the pathophysiology of HD and shows that defining circuitopathies of neuropsychiatric features plays an important role in understanding the disease.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Psychiatric (eg. Depression, Anxiety, Schizophrenia)
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling 2
Keywords:
Computational Neuroscience
Degenerative Disease
Movement Disorder
Psychiatric Disorders
Other - Depression; Effective Connectivity; Dynamic Causal Modelling; Default Mode Network
1|2Indicates the priority used for review
Provide references using author date format
Epping, E.A., et al. (2011). ‘Depression in the Early Stages of Huntington Disease’. Neurodegenerative Disease Management, vol. 1, no. 5, pp. 407–14
Friston, K.J., et al., (2014). ‘A DCM for Resting State fMRI’. NeuroImage, vol. 94, pp. 396–407
Friston, K.J., et al., (2016). ‘Bayesian Model Reduction and Empirical Bayes for Group (DCM) Studies’. NeuroImage vol. 128, pp. 413–31
Garcia-Gorro, C., et al. (2019). ‘Specific Patterns of Brain Alterations Underlie Distinct Clinical Profiles in Huntington’s Disease’. NeuroImage: Clinical, vol. 23, pp. 101900
Hannan, A.J. (2018). ‘Synaptopathy, Circuitopathy and the Computational Biology of Huntington’s Disease’. BMC Biology, vol. 16, no. 1, pp. 71
Klöppel, S., et al. (2015). ‘Compensation in Preclinical Huntington’s Disease: Evidence From the Track-On HD Study’. EBioMedicine, vol. 2, no. 10, pp. 1420–29
McColgan, P., et al. (2017). ‘Structural and Functional Brain Network Correlates of Depressive Symptoms in Premanifest Huntington’s Disease’. Human Brain Mapping, vol. 38, no. 6, pp. 2819–29
Razi, A., et al. (2015). ‘Construct Validation of a DCM for Resting State fMRI’. NeuroImage, vol. 106, pp. 1–14