Modeling the neurocognitive dynamics of language across the lifespan

Abstract

Healthy aging is associated with a heterogeneous decline across cognitive functions, typically observed between language comprehension and language production (LP). Examining resting-state fMRI and neuropsychological data from 628 healthy adults (age 18-88) from the CamCAN cohort, we performed state-of-the-art graph theoretical analysis to uncover the neural mechanisms underlying this variability. At the cognitive level, our findings suggest that LP is not an isolated function but is modulated throughout the lifespan by the extent of inter-cognitive synergy between semantic and domain-general processes. At the cerebral level, we show that default mode network (DMN) suppression coupled with fronto-parietal network (FPN) integration is the way for the brain to compensate for the effects of dedifferentiation at a minimal cost, efficiently mitigating the age-related decline in LP. Relatedly, reduced DMN suppression in midlife could compromise the ability to manage the cost of FPN integration. This may prompt older adults to adopt a more cost-efficient compensatory strategy that maintains global homeostasis at the expense of LP performances. Taken together, we propose that midlife represents a critical neurocognitive juncture that signifies the onset of LP decline, as older adults gradually lose control over semantic representations. We summarize our findings in a novel synergistic, economical, nonlinear, emergent, cognitive aging model, integrating connectomic and cognitive dimensions within a complex system perspective.

Keywords: SENECA; aging; cognition; connectomics; graph theory; language; reorganization.

FIGURE 2

System‐level topological dynamics across the lifespan. (a) Illustrates normalized efficiencies (y‐axis) as a function of age (x‐axis). Eloc = Local efficiency; Eglob = Global efficiency; Balance = dominance of global efficiency calculated as (Eglob − Eloc)/(Eglob + Eloc). (b, c) Evolution of the relative proportion of topological roles (y‐axis) with age (x‐axis). For the subsystem level (Panel c), changes with a tendential statistical significance are indicated by a star next to the label. RS NET (subsystems of the language connectome at rest); RS NET1 (associative); RS NET2 (sensorimotor); RS NET3 (bottom‐up attentional); RS NET4 (top‐down control‐executive). Please refer to Figure 1 for the composition of each RS NET. Connector (high integration/high segregation); provincial (low integration/high segregation); satellite (high integration/low segregation); peripheral (low integration/low segregation).

FIGURE 3

Probabilistic topological model of age‐related integration. Two dynamics of integration across the lifespan are proposed: (i) “energy‐costly” dynamic based on long‐range connections between subsystems, highlighting the flexible DMN‐FPN coupling in younger adulthood, and (ii) “energy‐efficient” dynamic based on short‐range connections within subsystems, highlighting a less flexible DMN‐FPN coupling in older adulthood. RS NET (subsystems of the language connectome at rest); RS NET1 (associative); RS NET2 (sensorimotor); RS NET3 (bottom‐up attentional); RS NET4 (top‐down control‐executive). Please refer to Figure 1 for the composition of each RS NET. Connector (high integration/high segregation); provincial (low integration/high segregation); satellite (high integration/low segregation); peripheral (low integration/low segregation). Default mode network (DMN), fronto‐parietal network (FPN), cingulo‐opercular network (CON), and sensorimotor network (SMN). Labels under brain illustrations are the names of the regions following the labeling proposed by Glasser et al. (2016). Brain visualization was done with the package ggseg in R (Mowinckel & Vidal‐Piñeiro, 2020) and projected on a multimodal cortical (HCP_MMP1.0; Glasser et al., 2016) and subcortical parcellation (Fischl et al., 2002).

FIGURE 4

Age‐related neurocognitive dynamics of language. (a, b) The age‐related trajectory of the cognitive variate and the corresponding neural mechanisms according to the results. The bar above the x‐axis reports age points with a significant second‐order derivative for most trajectories, reflecting a neurocognitive transition in midlife. (c, d) The structure coefficients with the cognitive variates, the correlations (cognitive variable‐cognitive variate), or cross‐correlations (brain mechanism‐cognitive variate). The red star indicates that naming and DMN suppression are highly covariant with each variate, thus showing that DMN suppression underpins naming performances during healthy aging.

FIGURE 1

Illustration of the modular organization of the language connectome at rest. Each RS NET is a subsystem obtained from the consensus clustering procedure (see Section 2.4.2). An additional three‐region module (black) associated with the ventral multi‐modal (VMM) network was also identified but not considered for analysis due to its instability as a stand‐alone module across the lifespan. LH (left hemisphere); RH (right hemisphere); RS NET1 (40 regions, red), RS NET2 (34 regions, orange), RS NET3 (32 regions, yellow), RS NET4 (22 regions, blue). Default mode network (DMN), fronto‐parietal network (FPN), cingulo‐opercular network (CON), and sensorimotor network (SMN). Brain visualization was done with the package ggseg in R (Mowinckel & Vidal‐Piñeiro, 2020) and projected on a multimodal cortical (HCP_MMP1.0; Glasser et al., 2016) and subcortical parcellation (Fischl et al., 2002). For details, please refer to Figures S1 and S2 in Appendix S2, and Appendix S3.