Late-onset depression (LOD) is featured by disrupted cognitive performance, which is refractory to conventional treatments and increases the risk of dementia. Aberrant functional connectivity among various brain regions has been reported in LOD, but their abnormal patterns of functional network connectivity remain unclear in LOD.

A total of 82 LOD and 101 healthy older adults (HOA) accepted functional magnetic resonance imaging scanning and a battery of neuropsychological tests. Static functional network connectivity (sFNC) and dynamic functional network connectivity (dFNC) were analyzed using independent component analysis, with dFNC assessed via a sliding window approach. Both sFNC and dFNC contributions were classified using a support vector machine.

LOD exhibited decreased sFNC among the default mode network (DMN), salience network (SN), sensorimotor network (SMN), and language network (LAN), along with reduced dFNC of DMN-SN and SN-SMN. The sFNC of SMN-LAN and dFNC of DMN-SN contributed the most in differentiating LOD and HOA by support vector machine. Additionally, abnormal sFNC of DMN-SN and DMN-SMN both correlated with working memory, with DMN-SMN mediating the relationship between depression and working memory. The dFNC of SN-SMN was associated with depressive severity and multiple domains of cognition, and mediated the impact of depression on memory and semantic function.

This study displayed the abnormal connectivity among DMN, SN, and SMN that involved the relationship between depression and cognition in LOD, which might reveal mutual biomarkers between depression and cognitive decline in LOD.

Previous studies have analyzed brain functional connectivity to reveal the neural physiopathology of bipolar disorder (BD) and major depressive disorder (MDD) based on the triple-network model [involving the salience network, default mode network (DMN), and central executive network (CEN)]. However, most studies assumed that the brain intrinsic fluctuations throughout the entire scan are static. Thus, we aimed to reveal the dynamic functional network connectivity (dFNC) in the triple networks of BD and MDD.

We collected resting state fMRI data from 51 unmedicated depressed BD II patients, 51 unmedicated depressed MDD patients, and 52 healthy controls. We analyzed the dFNC by using an independent component analysis, sliding window correlation and k-means clustering, and used the parameters of dFNC state properties and dFNC variability for group comparisons.

The dFNC within the triple networks could be clustered into four configuration states, three of them showing dense connections (States 1, 2, and 4) and the other one showing sparse connections (State 3). Both BD and MDD patients spent more time in State 3 and showed decreased dFNC variability between posterior DMN and right CEN (rCEN) compared with controls. The MDD patients showed specific decreased dFNC variability between anterior DMN and rCEN compared with controls.

This study revealed more common but less specific dFNC alterations within the triple networks in unmedicated depressed BD II and MDD patients, which indicated their decreased information processing and communication ability and may help us to understand their abnormal affective and cognitive functions clinically.