APOE ɛ4 and Insulin Resistance Influence Path-Integration-Based Navigation through Distinct Large-Scale Network Mechanisms

Abstract

Path integration (PI), which supports navigation without external spatial cues, is facilitated by grid cells in the entorhinal cortex. These cells are often impaired in individuals at risk for Alzheimer's disease (AD). However, other brain systems can compensate for this impairment, especially when spatial cues are available. From a graph-theoretical perspective, this compensatory mechanism might manifest through changes in network segregation, indicating shifts in distinct functional roles among specialized brain regions. This study explored whether similar compensatory mechanisms are active in APOE ε4 carriers and individuals with elevated insulin resistance, both susceptible to entorhinal cortex dysfunction. We applied a graph-theoretical segregation index to resting-state fMRI data from two cohorts (aged 50-75) to assess PI performance across virtual environments. Although insulin resistance did not directly impair PI performance, individuals with higher insulin resistance demonstrated better PI with less segregated brain networks, regardless of spatial cue availability. In contrast, the APOE effect was cue-dependent: ε4 heterozygotes outperformed ε3 homozygotes in the presence of local landmarks, linked to increased sensorimotor network segregation. When spatial cues were absent, ε4 carriers exhibited reduced PI performance due to lower segregation in the secondary visual network. Controlling cortical thickness and intracortical myelin variability mitigated these APOE effects on PI, with no similar adjustment made for insulin resistance. Our findings suggest that ε4 carriers depend on cortical integrity and spatial landmarks for successful navigation, while insulin-resistant individuals may rely on less efficient neural mechanisms for processing PI. These results highlight the importance of targeting insulin resistance to prevent cognitive decline, particularly in aging navigation and spatial cognition.

Figure 1.

Differences in functional network segregation between the two cohorts. Significant differences between the two cohorts for each functional network in the Gordon and Ji parcellation scheme. The shaded blue line indicates the p value of 0.05. The solid red line indicates the p value of the LME model comparing network segregation between the S_longAG and S_shortAG.

Figure 2.

Differences in cortical thickness between the two cohorts. Vertex-wise significant differences in cortical thickness between the two cohorts after adjusting by age, gender, education years, BMI, and APOE genotype. Statistical results are summarized in Table 2.

Figure 3.

Differences in intracortical myelin content between the two cohorts. Vertex-wise significant differences in intracortical myelin content derived from the T1w/T2w ratio between the two cohorts after adjusting by age, gender, education years, BMI, and APOE genotype. Statistical results are summarized in Table 2.

Figure 4.

Influence of APOE genotype and HOMA-IR on PI Performance across different subtasks. (A) Scatterplots depicting the relationship between HOMA-IR in the S_longAG (N = 53) and drop errors across different subtasks. Significant associations were observed when combining subtasks, but not individually. (B) Scatterplots showing the relationship between HOMA-IR in the S_shortAG (N = 49) and drop errors across different subtasks. No significant associations were found either when subtasks were combined or individually. The shaded area around the regression slopes in (A) and (B) represents the 95% confidence interval. (C) Performance of S_longAG participants (which is inversely related to drop error) was notably enhanced in APOE ɛ4 carriers when supportive spatial cues were available, i.e. in the LPI subtask. (D) Performance of S_shortAG participants showed no significant impact of APOE genotype in either the PPI or LPI subtask.

Figure 5.

Association between network segregation and PI as influenced by HOMA-IR. (A) Scatterplots depicting the relationship between sensorimotor network segregation in the Gordon parcellation scheme and drop error among individuals of the S_longAG (N = 53) with varying HOMA-IR levels. (B) Similar analysis as in (A) but for the sensorimotor network in the Ji parcellation scheme. (C) Similar analysis as in (A) but for the associative network in the Gordon parcellation scheme. The shaded area around the regression slopes represents the 95% confidence interval. Note that in all cases, the relationship was significant for individuals with higher HOMA-IR values.

Figure 6.

Association between sensorimotor network segregation and PI across the two cohorts. (A) Scatterplots illustrating the relationship between sensorimotor network segregation in the Gordon parcellation scheme and drop error in S_longAG for APOE ε4 carriers (n = 15) and non-carriers (n = 38). (B) Similar analysis as in (A) but for S_shortAG (ε4 carriers = 22; non-carriers = 27). The shaded area around the regression slope represents the 95% confidence interval. Note that the relationship was significant only for ε4 carriers.

Figure 7.

Impact of APOE genotype and HOMA-IR on cortical thickness. (A) Differences in cortical thickness between APOE ɛ4 carriers (ɛ4⁺) and non-carriers (ɛ4^-). Regions in red indicate cortical thinning and regions in blue cortical thickening in APOE ɛ4 carriers compared to noncarriers. (B) Association between cortical thickness and HOMA-IR. Regions in red indicate a positive association, while regions in blue indicate a negative association. Statistical results are summarized in Table 9.

Figure 8.

Impact of HOMA-IR and APOE genotype on intracortical myelin. (A) Differences in intracortical myelin, as derived from the T1w/T2w ratio, between APOE ɛ4 carriers (ɛ4⁺) and non-carriers (ɛ4^-). Regions in red indicate decreased intracortical myelin in APOE ɛ4 carriers compared to noncarriers. (B) Association between intracortical myelin content and HOMA-IR. Regions in red indicate a positive association with HOMA-IR. Statistical results are summarized in Table 10.