Disrupted modular brain dynamics reflect cognitive dysfunction in Alzheimer's disease
Highlights
► Network analysis applied to MEG data to study functional sub-networks (modules). ► In Alzheimer's disease, altered modular organization relates to cognitive symptoms. ► Intermodular connectivity is damaged most, parietal region has highest local damage.
Introduction
A theoretical framework to interpret the rapidly increasing amount of experimental data describing the complex organization of the human brain is highly desired. In recent years, graph theory has emerged as a promising candidate for this purpose (Bullmore et al., 2009, Rubinov and Sporns, 2010, Stam, 2010a, Stam, 2010b). Graph theory investigates the principles of network architecture, and the relation between network structure and function (Barabasi and Albert, 1999, Newman, 2010, Sporns, 2010, Watts and Strogatz, 1998). The application of graph theoretical analysis to neuroscientific data has revealed important organizational brain features such as an efficient ‘small-world’ architecture (combining good global and local connectivity) and the existence of highly connected network regions, called hubs (Achard et al., 2006, Eguiluz et al., 2005, He et al., 2008, Salvador et al., 2005, Stam et al., 2009, van den Heuvel and Hulshoff Pol, 2010). Changes in brain network topology have been related to normal cognitive development and aging as well as to a wide range of brain diseases, implying a close relation between connectivity and cognitive status (Achard and Bullmore, 2007, Bullmore and Sporns, 2009, Stam and Reijneveld, 2007).
In the most prevalent type of dementia, Alzheimer's disease (AD), cognitive functions that depend strongly on communication between different brain areas are particularly disturbed, and it has therefore been characterized as a ‘disconnection syndrome’ (Delbeuck et al., 2003, Geschwind, 1965). Graph theoretical studies of AD patient data have consistently revealed perturbations of brain network organization (de Haan et al., 2009, He et al., 2008, He et al., 2009, Stam and Reijneveld, 2007, Stam et al., 2009, Supekar et al., 2008). Interestingly, highly connected hub regions (e.g. the posterior cingulate gyrus and precuneus) seem most susceptible to AD pathology, which consists of amyloid deposition, hypometabolism and atrophy (Buckner et al., 2005, Celone et al., 2006, Greicius et al., 2004, Sperling et al., 2009). What causes this hub vulnerability in AD is unclear, but a more detailed description and understanding of hubs, or network clustering in general, could provide further clues.
A related network characteristic dealing with clustering is modularity, which expresses the extent to which networks can be decomposed into smaller functional sub-groups or modules (Boccaletti et al., 2006, Guimerà and Amaral, 2005, Newman, 2006, Newman and Girvan, 2004). Network nodes belonging to the same module have a higher level of inter-connectivity than with the rest of the network. In the brain, a high level of structural or functional connectivity among a group of regions implies a collective function or goal (Hilgetag et al., 2000, Salvador et al., 2005, Varela et al., 2001). Therefore, large-scale modular organization might be an appropriate level to examine cognitive processing and its impairment in brain disease. In this MEG study, we focus on functional modularity: the description of distinct sub-networks with intensive dynamical interaction, as expressed by levels of neuronal synchronization.
Theoretically, there are several advantages of a modular brain network structure. It offers an elegant solution for balancing the opposing demands that are placed on many dynamical systems: a high level of local specialization, while maintaining tight global integration (Sporns et al., 2004). In a modular network, hubs can have different roles; connector hubs form bridges between different modules, while provincial hubs are central nodes within modules. Graph theoretical measures that quantify inter- and intra-modular connectivity and are able to classify (hub) nodes accordingly have been developed (Guimerà and Amaral, 2005) and incorporated in the present study.
Using graph theoretical methods, several previous studies have demonstrated the presence of modular organization in the brain (Chen et al., 2008, Hagmann et al., 2008, Hilgetag et al., 2000, Kaiser et al., 2007, Leise, 1990). Moreover, modularity seems to develop during infancy and to degrade with age, suggesting a relation with cognitive abilities (Fair et al., 2009, Fan et al., 2011, Ferrarini et al., 2009, Meunier et al., 2009a, Meunier et al., 2009b, Schwarz et al., 2008, van den Heuvel and Hulshoff Pol, 2010). Consequently, the progressive impairment of specific cognitive domains in AD might well be reflected by changes in functional modularity.
In this study we explore functional modularity in resting-state MEG data of AD patients and healthy controls using a well-known graph theoretical modularity algorithm (Newman and Girvan, 2004). Our main aim is to examine whether and to what extent modular organization of spontaneous brain activity changes in AD, and if these changes are related to cognitive performance. Our hypothesis is that cognitive impairment in AD will be primarily reflected by impaired communication between functional modules, based on the notions that cognition requires intensive distributed processing and that vulnerable hub regions in AD are mainly located in association cortex areas (that integrate information from multiple modalities). In network terms, we expect AD to be a ‘connector hub disease’.
Section snippets
Patients and controls
The study involved 18 patients with a diagnosis of probable AD according to the NINCDS-ADRDA criteria (McKhann et al., 1984) and 18 healthy controls who were all recruited from the Alzheimer Center of the VU University Medical Center. Controls were often spouses of the patients. AD patients were assessed according to a standard clinical protocol, which involved history taking, physical and neurological examination, an interview with a spouse or close family member, blood tests, MRI of the brain
Modularity — descriptive results
To get a first impression of modular organization, individual network modules were visualized. Comparing several different resting-state MEG epochs of the same person, modular structure was generally consistent. Often, three or four strongly clustered frontal or parietal modules were found, along with several weaker temporal and occipital ones. Modules were usually localized clusters of adjacent cortical areas, but also showed long-distance fragments. Inter-hemispheric modules were a frequent
Discussion
The main message of this study is that the modular organization of large-scale spontaneous brain activity networks is disrupted in AD. Graph theoretical modularity analysis demonstrates weakening links within and, especially, between functional modules, correlating with cognitive dysfunction. Moreover, the vulnerability of the parietal region in AD is confirmed by regional analyses. In the following paragraphs we will relate our findings to current literature and discuss methodological issues.
Conclusion
It becomes more and more evident that disruption of structural and functional brain connectivity plays a pivotal role in the onset of dementia (Stam, 2010a, Stam, 2010b). Graph theory allows us to go beyond classifying AD as a disconnection syndrome, providing more detail and meaning. Functional modules are theoretically plausible representations of cognitive (sub-)processes, and therefore modularity analysis of MEG data seems a method with an appropriate spatiotemporal resolution to examine
Acknowledgments
The authors thank Nicole Sistermans, Ellemarije Altena, Annelies van der Vlies and Sofie Boom for neuropsychological assessments, and Karin Plugge and Ndedi Sijsma for performing the MEG recordings.
References (81)
- et al.
Complex networks: structure and dynamics
Phys. Rep.
(2006) - et al.
Generic aspects of complexity in brain imaging data and other biological systems
Neuroimage
(2009) - et al.
fMRI resting state networks define distinct modes of long-distance interactions in the human brain
Neuroimage
(2006) - et al.
Brain anatomical networks in early human brain development
Neuroimage
(2011) A mechanism for cognitive dynamics: neuronal communication through neuronal coherence
Trends Cogn. Sci.
(2005)- et al.
Studying the human brain anatomical network via diffusion-weighted MRI and graph theory
Neuroimage
(2008) EEG dynamics in patients with Alzheimer's disease
Clin. Neurophysiol.
(2004)Modular construction of nervous systems: a basic principle of design for invertebrates and vertebrates
Brain Res. Brain Res. Rev.
(1990)- et al.
Age-related changes in modular organization of human brain functional networks
Neuroimage
(2009) - et al.
Synchronization likelihood with explicit time-frequency priors
Neuroimage
(2006)