*5.2. rs-fMRI Studies Associated with the SN in LLD*

A disrupted standard pattern of SN connectivity is suggested to be one of the key traits of the pathogenesis of depression, particularly in the insula and amygdala [210]. Elevated connectivity between the insula and DMN was enhanced in MDD patients, which may hinder the above standard pattern [91]. The FC from the amygdala, another important SN node, to the hippocampus was decreased in adolescents with depression and at a high risk of depression [211,212]. Additionally, seed-based analysis in younger adults using the amygdala as a seed region was positively associated with increased amygdala FC with DMN nodes and long-term negative emotions [213]. One study that addressed apathy in LLD patients found that LLD patients with apathy exhibit increased FC between the SN and DMN compared with nonapathetic elders with depression [77]. Overall, these results may suggest that increased FC between the SN and DMN may predispose individuals to depression and is further correlated with vegetative symptoms in LLD [186]. However, inconsistent results for decreased FC between the amygdala and precuneus in depressed patients compared with controls have been reported [214].

Network analysis reported that elders with LLD also demonstrate a decreased negative FC between the SN and ECN compared to nondepressed age-matched controls [39]. Another study that compared correlation patterns among significant brain networks in LLD patients compared to nondepressed elderly controls reported dissociation patterns among the ECN/SN, and DMN observed in controls [215]. These results represent a failure of internetwork cohesiveness in LLD [185]. Moreover, decreased negative FC between the ECN and the SN was associated with cognitive impairment and severity of depression symptoms in LLD patients [39]. In addition, a worse treatment response to antidepressants was also associated with this disrupted standard SN pattern [216]. We presented key findings associated with LLD in Table 5 and characteristics of main rs-fMRI studies in Supplementary Table S6.

**Table 5.** Summary of key findings of rs-fMRI studies associated with the salience network (SN) in late-life depression (LLD) patients included in the review.


Abbreviations: DMN, default mode network; SN, salience network.

*5.3. rs-fMRI Studies Associated with SN in AD and MCI*

SN connectivity has increasingly gained attention from researchers who address neurodegenerative disease [133]. Although intensified SN connectivity was observed in AD patients compared to healthy controls in ICA studies [130,170], another ICA study in AD patients found contradictory evidence of a decrease in dorsal SN [121]. This increased SN connectivity has been consistently reported in cognitively normal individuals with elevated amyloid levels [166,217], Apo ε4 carriers [156,218], and MCI patients [74]. Moreover, studies that have addressed both amyloid and tau within the DMN and SN reported interesting results, with increased connectivity in the SN and DMN in individuals with elevated amyloid but little evidence of tau, but decreased connectivity in the SN and DMN in individuals with both elevated tau and amyloid levels [219]. These findings highlight the point that SN connectivity changes occur in preclinical dementia, and SN connectivity may change with disease progression. We presented key findings associated with AD and MCI in Table 6 and characteristics of main rs-fMRI studies in Supplementary Table S7.

**Table 6.** Summary of key findings of main rs-fMRI studies associated with the salience network (SN) in Alzheimer's disease (AD) and mild cognitive impairment (MCI) patients included in the review.


Abbreviations: SN, salience network; AD, Alzheimer's disease.

#### **6. Conclusions**

Alteration in brain networks during the resting state contributes to the symptoms and progression of LLD and AD. Above, we described LLD and AD, focusing on key networks known to be necessary for the network-level description of these two diseases: the DMN, ECN, and SN (Figure 1). A growing body of literature suggests an opposite direction for overall DMN alterations in LLD and AD, with increased connectivity of the DMN in LLD but decreased DMN connectivity in AD. However, the dissociation between the anterior DMN and posterior DMN provides insight into the link between depression and dementia. In the early stage of AD, the alteration in the DMN is different between its anterior and posterior subdivisions, with increased anterior DMN connectivity, and decreased posterior DMN connectivity [120]. Similar dissociation patterns were also observed in individuals with depression, and this increased anterior DMN persists after antidepressant treatments [99]. Additionally, a posterior DMN connectivity reduction was observed in individuals with LLD + MCI compared to LLD only [46,188] and also in LLD patients with an inadequate response to treatment [101]. Additionally, the PCC, the hub of the posterior DMN, is a marker of very early AD progression, as consistently seen with T1-weighted imaging, postmortems, and PET studies [220–222]. Although this association between the dissociation of DMN connectivity and AD and LLD remains to be explored, severe depression may induce the clinical manifestation of cognitive impairment or the onset of eventual cognitive decline, a signal of intrinsic network dysfunction.

Regarding the ECN, both AD and LLD exhibit disrupted ECN connectivity. As discussed, executive dysfunction associated with disrupted ECN connectivity seems to be related to the clinical prognosis of LLD with poor and slow antidepressant responses and a high relapse rate. The findings that the degree of ECN disruption is associated with cognitive decline 24 months after MCI is also covered above. With the hypothesis that depression precedes cognitive decline or induces cognitive decline [223], these results suggest the possibility that the ECN is a target that can modify the impact of LLD on cognitive declines. A noninvasive treatment is being conducted with the ECN as a target [224].

Another interesting issue seen in rs-fMRI studies is associated with the pathogenic process. Several studies using rs-fMRI associated with AD, tau, and amyloid pathology consistently reported that the spreading of these pathologies throughout the brain correlates to brain network disruption, as discussed in this review. Because the DMN, ECN, and SN are multimodal networks that are metabolically expensive and display high rates of cerebral blood flow, aerobic glycolysis, and oxidative glucose metabolism [225], these networks may be vulnerable to AD-associated pathogenic processes. Although spatial deposition patterns have been not convergent, there has been a recent observation that tau and amyloid plaques overlap with brain tissue loss in hub regions of these discussed

brain networks [226]. Another review also points to this correlation and suggests that AD-associated pathophysiological processes may explain changes in these networks [133].

**Figure 1.** The figure presents the aberrant functional connectivity between three networks in AD and LLD and its clinical implication. Blue line and red line indicate the decreased functional connectivity and the increased functional connectivity compared with healthy control, respectively. Decreased functional connectivity of the executive control network was commonly observed in AD and LLD [192,197]. In contrast, the functional connectivity of the salience network and the default mode network were differently altered. The default mode network connectivity decreased in AD and increased in LLD [93,127], and the salience network increased in AD and decreased in LLD [130,215]. However, dissociated functional connectivity pattern in DMN, increased in the anterior DMN and decreased in the posterior DMN, was commonly observed in both AD and LLD [96,166]. This dissociation reflects treatment effects in LLD and amyloid/tau pathology in AD [100,165]. This similarity of dissociation seems to be a possible mechanism of association between LLD and AD highlighted in epidemiological studies.

Despite the consistent findings across studies, several critical knowledge gaps remain. The lack of standardized protocol for addressing the brain using rs-fMRI has not been adequately addressed. Regarding preprocessing steps of fMRI for dealing with noise, preprocessing steps for rs-fMRI data have evolved to be more diverse than preprocessing for task-based fMRI data. With the diversity of statistical approaches applied to the purified data, these nonstandardized various methods make comparisons across studies extremely difficult. Even if the same terms are used to describe results such as network strength or connectivity, one method's results cannot be compared well with the results of studies using other technologies.

Additionally, our literature did not include task-based fMRI studies in AD and LDD, which clearly expressed the need for additional research. Compared to rs-fMRI studies, a task-based fMRI study is a relatively conventional approach and is challenging to perform due to the needs of involving tasks. Nevertheless, preprocessing steps and statistical methods for task-based fMRI have been more standardized than those for rs-fMRI and have less influenced results, where an external behavioral standard is available. Because taskbased fMRI studies are more clinically interpretable, future studies that include tailored tasks concerning specific cognitive, emotional, and social functions would expand our knowledge of AD and LLD.

Recent studies using directed graph theory or combining multiple imaging tools have presented promising results in the field of diagnosis and prediction [227,228]. Therefore, future studies combining multimodal imaging tools such as PET, MRI, and fMRI in AD and LLD patients samples with special considerations such as age, sex, age of onset, treatment outcomes, the severity of illness, and cognitive impairment would help us understand the fundamental functional pathological changes in AD and LLD. Longitudinal studies that include various treatment tools would also help uncover the association between depression and AD-associated pathophysiological processes. Standardized protocols in fMRI data collection and analysis would be helpful to reduce heterogeneity across these physiological states.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2227-905 9/9/1/82/s1.

**Author Contributions:** J.K. reviewed and summarized articles, prepared figures, and wrote the manuscript. Y.-K.K. conceived the structure of the paper, supervised the review, organized the figures, and wrote the manuscript. All authors contributed to the critical reading and writing of the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** No new data were created or analyzed in this study. Data sharing is not applicable to this article.

**Conflicts of Interest:** The author declares no conflict of interest.
