Default mode cortex supports information integration

Recent work leveraging the intrinsic connectivity of the brain has shed light on the topographical organization of heteromodal default mode cortex (Braga et al., 2013; Buckner and Krienen, 2013; Leech et al., 2012; Margulies et al., 2016). By applying decomposition techniques to brain connectivity, Margulies and colleagues (2016) characterized a principal gradient of cortical organization (Figure 1.5.A) which is anchored at one end by unimodal regions involved in perception and action, and at the other end by transmodal regions including angular gyrus, medial PFC, anterior and posterior cingulate, and anterior temporal lobe. The observation that DMN nodes at the top of the gradient are maximally distant from unimodal cortices (Figure 1.5.B) supports the proposed role of this network in functions that require integration of multiple features (Margulies et al., 2016). The topographic separation of heteromodal default cortex from primary systems suggests that increasingly complex and abstract functions might be formed along the gradient (Figure 1.5.C), where the input from unimodal features is progressively reduced (Buckner and Krienen, 2013; Mesulam, 1998; Plaut, 2002; Schapiro et al., 2013). In this view, complex relationships might be more easily captured where there is greater separation between unimodal “spokes”

and association regions in default mode network (Lambon Ralph et al., 2016; Margulies et al., 2016).

Distance might also allow the brain to support forms of cognition that require to focus on previously encoded knowledge, as opposed to information in the external environment (i.e. memory; Murphy et al., 2019).

38 Figure 1.5. A Principal gradient of connectivity projected on a template surface (Margulies et al., 2016). The gradient is a continuous map with values ranging from 0 (in dark blue) to 100 (red). B Distance of each point on the cortical surface from seven seeds placed within the default mode network (DMN). Core DMN nodes (red) are located at maximal geodesic distance from landmarks of unimodal function such as the central sulcus (motor cortex) and the calcarine sulcus (visual cortex). This topographical organization divides the cortical surface into distinct “zones”

(separated by white lines). C Results of a Neurosynth meta-analysis in regions of interest along the gradient, showing increasing abstraction of functions towards the heteromodal end. Reproduced with permission from Margulies et al., 2016.

The proposed topographical organization of cortex with DMN regions at the top of the abstraction hierarchy is consistent with other connectivity studies showing “echoes” or traces of other large-scale networks in default mode cortex (Braga et al., 2013; Leech et al., 2012; Turnbull et al., 2019a). Using decompositions of resting-state fMRI data, Braga and colleagues (2013) were able to show that core nodes of the default mode network, including posterior and anterior cingulate cortex, angular gyrus and middle temporal gyrus, contained neural signals of other established large scale networks. In contrast, unimodal regions presented few traces of other networks and showed a more unitary pattern of intrinsic activity (see also Leech et al., 2012). These results were taken as evidence that DMN regions mediate the communication between different networks and therefore might play a role in the dynamic integration of information (for review see Braga and Leech, 2015). Interestingly, decomposition of the neural signal in the right supramarginal gyrus/AG node has revealed a local topographical organization, such that echoes of multiple large-scale networks were present in spatially adjacent but separable regions (Braga et al.,

39 2013). In addition to supporting a special role of the parietal default mode cortex in global integration, this observation is relevant to understanding why this region is implicated in multiple tasks. The activation of AG could reflect the summation of different neural signals, but in certain circumstances it could reflect the segregation between different components. Congruently with this view, a recent study by Kernbach et al.

(2018) revealed that patterns of functional coupling within DMN sub-regions reflected distributed connectivity of large-scale networks, suggesting that a key feature of DMN is to mediate the functional communication between networks. Interestingly, among the candidate regions in their DMN parcellation, the right temporo-parietal junction was found to be the one that explained the most variance in the interplay of DMN and other brain-networks. Convergent evidence for a role of AG in mediating the cross-talk between networks comes from a graph theory study by Xu et al. (2016) showing that AG acts as an

“overall connector hub” with strong connections to all modules in the graph.

In line with the proposed role of DMN in information integration, Vatansever et al. (2015b) found evidence for the functional relevance of DMN in goal-directed tasks. Using a working memory paradigm in fMRI and graph theoretical measures, the study revealed that DMN increased interactions with other large-scale networks under higher task demands (1 back vs. 0 back), suggesting a role of this network in the flexible integration of elements from memory. More recently, Evans et al. (2020) found that improved performance in a semantic summation task requiring participants to detect weak overlapping patterns of semantic association was related to increased connectivity between DMN and the control network. Similar patterns of connectivity between DMN and other large scale brain systems have been shown to support response inhibition during idea production (Beaty et al., 2017) and creative thinking abilities (Beaty et al., 2018, 2015; for review see Beaty et al., 2016). Moreover, default mode nodes including AG are recruited during the generation of metaphors (Benedek et al., 2014) and chain free associations (Marron et al., 2018), while differences in the DMN intrinsic connectivity are predictive of individual differences in creative thinking abilities (Beaty et al., 2014; Chen et al., 2015; Jauk et al., 2015; Jung et al., 2016). A fundamental component of creativity is the ability to flexibly re-combined stored concepts to produce novel ideas (Beaty et al., 2017). Consequentially, the DMN involvement in creative cognition might be similar to its contribution to other forms of cognition that rely on memory retrieval as opposed to information present in the environment (e.g. Margulies et al., 2016; Murphy et al., 2019). Moreover, recent task-based fMRI studies have suggested that semantic regions allied to DMN (including ATL and AG), support the combination of concepts into meaningful and more complex representations (Bemis and Pylkkänen, 2011; Price et al., 2015; Teige et al., 2019, 2018). Finally, DMN has been shown to support integration over long time-scales during narrative comprehension (Chang et al., 2020; Lerner et al., 2014,

40 2011; Simony et al., 2016). A study by Lerner et al. (2011) revealed a hierarchy in the temporal lobe responsible for accumulating information over increasing temporal scales. Within this cortical topography, auditory regions responded to fast-changing information reflecting momentary features of the input, adjacent areas in superior temporal gyrus combined information at the time-scale of single sentences, and heteromodal regions overlapping with DMN accumulated information across paragraphs. More recently, Chang et al. (2020) showed re-activation of the neural patterns associated with previously presented stories when the storylines converged and could be integrated.

Taken together, these findings point to a critical role of DMN in conceptual integration and are overall consistent with recent accounts of brain organization (Margulies et al., 2016). In line with the topographical architecture described by the principal gradient increasingly complex and abstract representations – key features of memory retrieval, creative thinking, and semantic integration – might emerge at the top end of the cortical hierarchy in regions overlapping with DMN, where the influence of sensory-motor system is minimal.