Top-down attention and Alzheimer's pathology affect cortical selectivity during learning, influencing episodic memory in older adults

Abstract

Effective memory formation declines in human aging. Diminished neural selectivity-reduced differential responses to preferred versus nonpreferred stimuli-may contribute to memory decline, but its drivers remain unclear. We investigated the effects of top-down attention and preclinical Alzheimer's disease (AD) pathology on neural selectivity in 166 cognitively unimpaired older participants using functional magnetic resonance imaging during a word-face/word-place associative memory task. During learning, neural selectivity in place- and, to a lesser extent, face-selective regions was greater for subsequently remembered than forgotten events; positively scaled with variability in dorsal attention network activity, within and across individuals; and negatively related to AD pathology, evidenced by elevated plasma phosphorylated Tau₁₈₁ (pTau₁₈₁). Path analysis revealed that neural selectivity mediated the effects of age, attention, and pTau₁₈₁ on memory. These data reveal multiple pathways that contribute to memory differences among older adults-AD-independent reductions in top-down attention and AD-related pathology alter the precision of cortical representations of events during experience, with consequences for remembering.

Fig. 1.. Experimental design, group-level statistical maps, and ROIs.

(A) Structure of study (encoding) and test (retrieval) trials in the word-picture associative memory task. (B) Group-level category effects (face-related versus place-related neural activity) during word-picture study. (C) Group-level SMEs depicting differential activation on word-picture encoding trials for which the association was subsequently remembered (Rem) versus subsequently forgotten (Forg) at test. The colored brain areas in (B) and (C) represent significant results after family-wise error (FWE) and threshold-free cluster enhancement (TFCE) whole-brain correction. The correction was applied separately for the left hemisphere, right hemisphere, and subcortical regions, with a threshold of −log₁₀(α/N) = −log₁₀(0.05/3) = 1.7782. These significant regions were then mapped onto the t-statistical maps. (D) Predefined place- and face-selective ROIs. (E) Predefined frontoparietal DAN and ventral attention network (VAN) ROIs.

Fig. 2.. Leave-one-participant-out ROI delineation.

For a given participant, we first performed a group-level analysis without the participant to obtain a whole-brain category effect in standard MNI space. A significant category effect map was extracted after controlling for FWE with P < 0.05 using TFCE. Next, the resulting whole-brain map was intersected with the predefined functional ROIs, thus yielding study-specific and participant-independent ROIs. Last, we transformed the study-specific ROIs to the given participant’s brain in native space.

Fig. 3.. Neural selectivity relates to age, top-down attention, and early AD pathology.

(A) The age-related decrease in neural selectivity at encoding is greater on subsequently remembered versus forgotten trials. a.u., arbitrary units. (B) Neural activity decreased more with age for preferred category than nonpreferred category trials in (left) place-selective but not (right) face-selective regions (analyses restricted to subsequently remembered trials). Sex and years of education were included as nuisance variables. (C) Neural selectivity on remembered trials was associated with the SME in frontoparietal nodes of the DAN. (D) DAN SME was tightly related to face- and place-related activity on subsequently remembered trials in (left) place-selective and (right) face-selective regions. (E) Neural selectivity on remembered trials in (left) place-selective but not (right) face-selective regions significantly declined with plasma pTau₁₈₁. (F) Plasma pTau₁₈₁ negatively related to preferred neural activity (i.e., place-related activity, cool color) in place-selective regions. Age, sex, and years of education were included as nuisance variables. Rem_Face and Rem_Place, face-related and place-related neural activity on subsequently remembered trials.

Fig. 4.. Predictors on neural selectivity.

(A) Plasma pTau₁₈₁ and DAN SME partially mediated the negative relationship between age and neural selectivity on remembered trials in place-selective regions. Sex and years of education were included as nuisance variables. ^#P = 0.05, *P < 0.05, **P < 0.01, and ***P < 0.001. c′ = direct effect; c = total effect = a₁b₁ + a₂b₂ + c′. (B) Unique and (C) shared explained variances of predictors of neural selectivity. DAN, DAN SME (i.e., SME in DAN); AD, plasma pTau₁₈₁; EDUC, years of education.

Fig. 5.. Greater neural selectivity is related to better memory performance.

(A) Neural selectivity in place-selective regions predicted in-scanner associative memory (overall associative d′) on remembered trials. (B) Neural selectivity in place-selective regions on remembered trials predicts proportion of exemplar-specific recall on the postscan test. (C and D) Neural selectivity in place-selective regions on remembered trials predicts out-of-task performance on (C) the delayed recall score during neuropsychological testing and (D) mnemonic similarity task (MST), where the similarity slope on the y axis reflects the magnitude of the increase in performance as target-lure similarity moved from high to low. (E) Neural selectivity in place-selective regions on subsequently remembered trials was not significantly associated with executive function.

Fig. 6.. Pathways and predictors on memory performance (i.e., place associative d′).

(A) Structural equation modeling. Solid lines represent significant paths; dash-dotted lines represent nonsignificant paths. The figure shows standardized betas. ^#P = 0.05, *P < 0.05, **P < 0.01, and ***P < 0.001. c′ = direct effect; c = total effect = a₁b₁ + a₂b₂ + a₃b₃ + a₁d₂₁b₂ + a₃d₂₃b₂ + c′. (B) Unique and (C) shared explained variances of predictors of place associative d′. NS, neural selectivity.