Functional reorganization of the large-scale brain networks that support high-level cognition following brain damage in aphasia

• ¹ MIT, Department of Brain and Cognitive Sciences, United States
• ² Boston University, Department of Speech, Language and Hearing Sciences, United States
• ³ MGH, Department of Psychiatry, United States

Over the last decade, a number of large-scale networks in the human cortex that support high-level cognition have been identified. Here, we focus on two of these networks: the fronto-temporal language network (e.g., Fedorenko et al., 2010), and the fronto-parietal “multiple demand (MD)” network (e.g., Duncan, 2010). These two networks are clearly distinct from one another: first, their respective regions show distinct functional profiles, with language regions showing selective responses to language stimuli (Fedorenko et al., 2011; Monti et al., 2012) and MD regions showing domain-general responses to cognitive effort across a wide range of tasks (Duncan & Owen, 2001; Fedorenko et al., 2013). Second, during “rest” and cognitive processing, each network shows strong activity synchronization among its constituent regions, whereas regions across the two networks are not synchronized (Blank et al., 2014; Lee et al., 2012; Mantini et al., 2013). In the current study, we examined how these functional characteristics of the two networks were affected following aphasia-inducing strokes. In particular, we asked whether damage to the language network would alter the involvement of the MD network in linguistic processing, and whether such damage would alter the patterns of synchronization across the two networks. Four male individuals with aphasia (age: M=53), having suffered a single left MCA CVA, were scanned in fMRI on two paradigms that enable basic functional characterization of language and MD regions: (i) a language localizer task, where they passively read sentences and sequences of pseudowords (Fedorenko et al., 2010); and (ii) a spatial working memory task, where they had to remember fewer (easy) or more (hard) locations in a grid (Fedorenko et al., 2013). Language and MD regions were defined in each individual using the sentences > pseudowords contrast and the hard > easy contrast, respectively. Subjects were also scanned while listening to naturalistic stories. We found that language regions maintained their selectivity, showing strong responses to sentences (that were reliably stronger than responses to pseudowords) and little or no response during the spatial working memory task. In contrast, MD regions showed strong responses to the spatial working memory task, with a few regions responding more strongly during the hard condition than during the easy condition. These patterns replicate previous findings in healthy individuals (Fedorenko et al., 2011, 2012). However, unlike in healthy controls – where MD regions respond more strongly to pseudowords than to sentences (presumably because processing pseudowords is more demanding) – in individuals with aphasia a few MD regions responded more strongly to sentences than to pseudowords. These regions were located in the bilateral opercular inferior frontal gyrus, precentral gyrus and supplementary motor area. Nonetheless, language and MD regions remained overall dissociated during story comprehension: regions within each network were more synchronized with each other than with regions of the other network. These results suggest that parts of the domain-general MD network, a system that has been linked to problem solving and general fluid intelligence (e.g., Duncan, 2010; Woogar et al., 2010), may alter their involvement in language processing when the language system is compromised.