Elevated plasma GFAP levels in MCI link APOE ε4 allele with impaired gait speed

Abstract

The presence of at least one copy of the apolipoprotein ε4 allele (APOE ε4) is a known predictor of gait impairment risk among older adults. However, the mechanisms by which APOE ε4 affects gait performance remain unclear. This cross-sectional study aimed to reveal underlying pathological mechanisms linking APOE ε4 carriage to slow gait. This secondary analysis used baseline assessments from the J-MINT multicenter intervention trial, focusing on older adults with mild cognitive impairment. Gait speed was measured at baseline, with slow gait (SG) defined as speeds one standard deviation below the age- and sex-specific mean. APOE phenotype and plasma biomarkers related to Alzheimer’s disease (AD), including amyloid-β composite biomarker, phosphorylated Tau 181, neurofilament light, and glial fibrillary acidic protein (GFAP), were also measured. The analysis included 236 non-APOE ε4 carriers and 84 carriers of at least one APOE ε4. APOE ε4 carriers exhibited significantly slower gait speed than non-carriers (1.20 m/s [SD = 0.22] vs 1.26 m/s [SD = 0.23], p = 0.042). Significant interaction between APOE ε4 carriage and SG was observed only in plasma GFAP levels (F_{1, 312} = 7.17, p = 0.008), indicating that individuals with APOE ε4 and SG had significantly higher plasma GFAP levels. Elevated plasma GFAP levels fully mediated the association between APOE ε4 carriage and gait speed (partially standardized indirect effect = −0.059: −0.12 to −0.013]). No other AD-related biomarkers mediated this association. Our results suggest that APOE ε4–related gait changes may reflect AD pathology, as indicated by elevated GFAP levels, and could potentially accelerate dementia symptoms.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11357-025-01955-5.

Introduction

The apolipoprotein ε4 allele (APOE ε4) is the major genetic risk factor for late-onset Alzheimer’s disease (AD) and associated cognitive impairments [1, 2]. Moreover, APOE ε4 has also been linked to physical functioning, including gait performance [3]. For example, rapid motor decline, such as muscle weakness, has been observed among older adults carrying the APOE ε4 allele [4]. A cross-sectional study revealed that these individuals exhibit slower gait speeds than those without the allele [3]. Our previous study focusing on older adults with mild cognitive impairment (MCI) confirmed these findings, showing that at least one copy of the APOE ε4 allele is associated with future declines in gait performance, specifically in gait stride time variability and gait stride length variability [5]. As gait performance is closely linked to cognition [6, 7], it is reasonable to conclude that APOE ε4 is involved in both cognitive and gait functions. However, the underlying mechanism remains unclear.

One possible explanation for the association between APOE ε4 status and lower gait performance is AD-related pathological changes. AD pathology is known to disrupt shared brain networks involved in both cognition and motor function, including gait. APOE ε4 accelerates the accumulation, aggregation, and deposition of amyloid-β (Aβ) and tau in the brain [8], likely impairing physical function by compromising neural circuits essential for motor control and coordination [9–11]. Additionally, neurotoxicity, inflammation, and cerebrovascular damage may further compromise the structural and functional integrity of motor-related brain regions. APOE ε4 carriers also exhibit higher levels of plasma neurofilament light (NfL) levels, a protein biomarker for axonal damage in the brain associated with atrophy, cognitive decline, and other markers of AD [12]. NfL levels in the blood can indicate neuroaxonal damage and predict the progression of movement disorders [13, 14]. Furthermore, recent studies have identified glial fibrillary acidic protein (GFAP)—an intermediate filament protein and classical astrocyte marker—as a potentially more sensitive indicator of AD pathology. GFAP levels are elevated in APOE ε4 carriers [15] and serve as early markers of Aβ-related changes [16]. Importantly, increased GFAP has also been linked to postural instability and gait disturbances [17, 18]. Taken together, the associations between APOE ε4 and AD pathology, and between AD pathology and impaired gait, suggest that the relationship between APOE ε4 status and reduced gait performance is mediated by AD-related neurodegenerative processes.

Understanding the underlying mechanisms of the association between APOE ε4 and gait performance among older adults may provide a neurophysiological explanation for why gait serves as a surrogate marker for cognitive impairment [19–21], which is associated with both aging and neurodegeneration. This information is particularly relevant for older adults with MCI, who are progressing at an accelerated rate to AD. Thus, this study aimed to examine whether AD-related biomarkers mediate the association between APOE ε4 carriage and gait performance in older adults with MCI. We conducted a two-step analysis. First, we identified AD-related biomarkers affected by an interaction between APOE ε4 and slower gait speed. Then, we performed mediation analysis based on all significant interactions. Additionally, we assessed the involvement of vascular burden factors as secondary endpoints, because individuals who carry APOE ε4 have a higher vascular risk burden [22, 23]. Based on prior evidence that Aβ and GFAP, which are elevated in APOE ε4 carriers and involved in the early stages of AD, are associated with motor dysfunction, we hypothesized that these biomarkers mediate the link between APOE ε4 carriage and slow gait.

Materials and methods

Participants

This study is a secondary analysis of baseline assessments from the J-MINT trial, an 18-month, open-label, randomized, controlled, multicenter trial that targets modifiable risk factors for dementia among older adults with MCI. Detailed participant characteristics, including methodologies for diagnosis/classification, have been previously reported [24, 25]. This study involved five centers located in Aichi and Tokyo, Japan. General inclusion criteria were individuals aged 65–85 years at enrollment with MCI, as assessed using the National Center for Geriatrics and Gerontology Functional Assessment Tool (NCGG-FAT), a screening tool for older adults at high risk of dementia [26]. MCI was defined as age- and education-adjusted cognitive decline, with a standard deviation (SD) of 1.0 or more from the reference threshold for one or more cognitive domains: memory, attention, executive function, and processing speed, as measured by the NCGG-FAT [27, 28]. Exclusion criteria included restrictions on physical exercise and/or diet due to functional decline, a diagnosis of dementia, a Mini-Mental State Examination (MMSE) score below 24, and a care certificate for long-term care (activities of daily living-dependent).

Although 531 participants were enrolled in the J-MINT baseline assessment, this study analyzed data from the NCGG cohort (n = 361), as blood AD-related biomarkers such as Aβ were measured only in this cohort. Ethical approval was obtained from the Ethics Committee of Human Research at the National Center for Geriatrics and Gerontology (No. 1791) and the trial has been registered with the University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR) under registration number UMIN000038671. UMIN is approved by the WHO Primary Registry.

APOE phenotype and plasma AD-related biomarkers

Plasma was isolated from whole blood collected in 7-mL EDTA-2Na tubes (Venoject II; TERUMO, Tokyo, Japan). Within 5 min of blood collection, whole blood was centrifuged at 2273 g for 5 min at room temperature. Alternatively, the blood was temporarily stored on ice for up to 30 min before centrifugation. The plasma was then transferred to storage tubes (48 Jacket Tubes, 2.0 mL External-Type; FCR & Bio, Tokyo, Japan) as 250 µL aliquots and frozen in a −80 °C freezer within 4 h whenever possible.

The APOE phenotype was classified using a standardized isoelectric focusing (IEF) method for immunoblotting. Participants were assigned as either APOE ε4 carriers (E4/E2, E4/E3, and E4/E4) or non-APOE ε4 carriers (E2/E2, E3/E2, and E3/E3).

Aβ composite biomarker, phosphorylated Tau (p-Tau) 181, NfL, and GFAP were assessed as plasma AD-related biomarkers. A plasma Aβ composite biomarker, which was generated scores by combining plasma ratios of APP669–711/Aβ1–42 and Aβ1–40/Aβ1–42, was measured using an immunoprecipitation-mass spectrometry assay, to assess Aβ deposition in the brain. Specifically, the composite value was calculated by normalizing the Z-scores of APP669–711/Aβ1–42 and Aβ1–40/Aβ1–42, based on data from NCGG-FAT [29], using the following formula:

{[(5.241 × APP669-711/Aβ1–42) − 4.056] + [(0.232 × Aβ1–40/Aβ1–42)−5.733]}/2.

This composite biomarker has been validated across multiple cohorts and demonstrates high accuracy in detecting elevated Aβ burden and cognitive impairment, with a reference cut-off value of 0.376 [29, 30]. Plasma concentrations of other biomarkers were measured using the single-molecule array (Simoa™) platform [31, 32]. Although several definitions have been reported, 2.2 pg/mL and 35.02 pg/mL have been suggested as cut-off values for p-Tau 181 and NfL, respectively [33, 34]. A GFAP cut-off value for dementia, measured using the same method as in this study, has not yet been proposed; however, the reported mean level in AD patients is 404.7 pg/mL [35].

Gait performance

Using the short physical performance battery [36], we evaluated gait speed with a 2.4-m gait test, which has demonstrated good external validity compared to the long-distance walking test [37]. Participants were asked to walk over a 4.4 m distance at their usual pace, twice. After the first walk, they stopped, turned around in place, and returned along the same path to complete the second trial. The first and last 1 m were considered acceleration and deceleration phases and were excluded from measurement. We used the average of two trials to approximate usual walking speed and enhance measurement stability, rather than the faster trial, which may reflect best-effort performance. We then defined a conventional slow gait as being one standard deviation below the mean gait speed, specific to age and sex, based on previous studies [38]. Slow gait, as defined by this method, has been strongly associated with adverse health outcomes, including cognitive decline and falls.

Vascular burden factors

We measured glycated hemoglobin (HbA1c), high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, and total cholesterol as vascular burden factors. Furthermore, a vascular risk index was assessed using a validated 5-point scale of vascular burden, which includes hypertension, dyslipidemia, diabetes mellitus, obesity, and smoking history [39]. Obesity was defined as a body mass index ≥ 25 kg/m² based on Japanese population criteria. Scores ranged from 1 to 5, with one point added for each risk factor present.

Covariates

Participants were interviewed regarding relevant sociodemographic and clinical variables as covariates for the association between APOE ε4 and slow gait, including age, sex, medication use (≥ 5 medications), and number of comorbidities.

Data analysis

Descriptive statistics comparing non-APOE ε4 and APOE ε4 carriers were analyzed using chi-square tests and t-tests, focusing on clinical and gait-related variables. A two-factor mixed analysis of variance (ANOVA) was conducted on AD biomarkers with two between-subject factors—APOE phenotype (non-APOE ε4 carriers vs. APOE ε4 carriers) and gait performance (with vs. without slow gait)—to examine the interaction between APOE phenotype and gait performance on AD biomarkers. Analyses were performed using IBM SPSS Statistics (Version 24.0), with the significance level set at p < 0.05. For post hoc tests following ANOVAs, Bonferroni correction was applied to control for type 1 error.

Mediation analysis, adjusting for covariates, was performed to examine whether AD markers explain or mediate the association between APOE ε4 and slow gait, based on significant interactions identified in the ANOVAs. These analyses were also conducted for vascular burden factors. When a significant indirect effect of AD markers, excluding Aβ, was observed, additional analyses were performed to determine whether the effect was associated with AD pathology or vascular risks. This distinction is important because increases in tau, NfL, and GFAP may not always result from AD-related changes but can also be attributed to vascular risk burden [40, 41], which is frequently observed in individuals carrying APOE ε4 [22, 23]. Specifically, we examined whether the Aβ composite biomarker, a pathological hallmark of AD, or the vascular risk index mediated the association between APOE ε4 and the significant AD markers linked to slow gait. These mediation analyses were conducted using the PROCESS macro for SPSS by Preacher and Hayes (Version 3.5), which generated 95% confidence intervals for the indirect effects based on 5000 bootstrap samples. We further applied the Aroian test to assess indirect effects. In mediation analysis, full mediation is indicated when the beta weight decreases and the result is no longer significant, whereas partial mediation is indicated when the beta decreases but remains significant.

Results

Prevalence and characteristics of APOE ε4 carriers

In total, 41 participants, representing 11.4% of the original sample, were excluded from the analysis due to missing data for key variables, such as gait and AD biomarkers. Among the 320 older adults with complete baseline key variables, the distribution of APOE phenotypes was as follows: E2/E2, n = 1 (0.3%); E3/E2, n = 28 (8.8%); E3/E3, n = 207 (64.7%); E4/E2, n = 3 (0.9%); E4/E3, n = 72 (22.5%); and E4/E4, n = 9 (2.8%). In total, 236 participants were non-APOE ε4 carriers, and 84 were carriers of at least one APOE ε4 allele. Table 1 presents participant characteristics and highlights differences between non-APOE ε4 and APOE ε4 carriers. Participants with APOE ε4 had higher education levels, significantly slower gait speeds, and a higher prevalence of falls within a year compared to non-APOE ε4 carriers.

Table 1.

Characteristics of participants stratified by APOE ε4 carrier status

	Non-APOE ε4 carrier	APOE ε4 carrier	p-value
	n = 236	n = 84
Female, n (%)	104 (44.1)	27 (32.1)	0.056
Age, mean (SD)	73.9 (4.9)	74.1 (5.1)	0.727
Number of years of education, mean (SD)	12.3 (2.4)	13.1 (2.3)	0.008
Five plus medications, n (%)	65 (27.5)	26 (31.0)	0.552
Number of comorbidities, mean (SD)	1.53 (1.30)	1.74 (1.41)	0.209
SBP, mmHg, mean (SD)	139.2 (17.9)	142.1 (19.9)	0.212
DBP, mmHg, mean (SD)	84.4 (11.5)	85.0 (9.9)	0.701
BMI, mean (SD)	23.2 (3.6)	23.0 (2.9)	0.691
MMSE, mean (SD)	27.6 (1.8)	27.6 (1.8)	0.802
Gait speed, m/s, mean (SD)	1.26 (0.23)	1.20 (0.22)	0.042
Fall experience within a year, n (%)	54 (22.9)	31 (36.9)	0.012

SD standard deviation, SBP systolic blood pressure, DBP diastolic blood pressure, BMI body mass index, MMSE mini-mental state examination

AD biomarkers affected by an interaction between APOE ε4 and gait

Table 2 shows AD biomarker values based on APOE ε4 carriage and slow gait. According to our criteria, 48 older adults (15.0% of the sample) were identified as having slow gait. Their mean gait speed was 0.91 m/s (SD = 0.19), whereas those without a slow gait had a mean gait speed of 1.30 m/s (SD = 0.17). Two-way ANOVAs on the composite biomarker (i.e., levels of plasma Aβ) and p-tau 181 revealed a significant main effect of the APOE phenotype (F_{1, 312} = 27.6, p < 0.001 and F_{1, 312} = 4.5, p = 0.034, respectively), but neither the main effect of gait performance (F_{1, 312} = 0.49, p = 0.484 and F_{1, 312} = 0.38, p = 0.538) nor the interaction between the two factors showed significance (F_{1, 312} = 0.22, p = 0.637 and F_{1, 312} = 0.34, p = 0.562). Higher biomarker values were confirmed in APOE ε4 carriers. Conversely, a significant interaction was observed in plasma GFAP levels (F_{1, 312} = 7.17, p = 0.008) with significant main effects of APOE phenotype (F_{1, 312} = 17.8, p < 0.001) and gait performance (F_{1, 312} = 4.8, p = 0.029). Post hoc analysis results (Fig. 1A) indicated that APOE ε4 carriers with slow gait had significantly higher plasma GFAP levels than non-APOE ε4 carriers with slow gait (p < 0.001) and APOE ε4 carriers without slow gait (p = 0.004). No significant main effects of APOE phenotype (F_{1, 312} = 0.13, p = 0.709), gait performance (F_{1, 312} = 1.2, p = 0.267), and interaction (F_{1, 312} = 0.20, p = 0.888) were found for NfL levels.

Table 2.

ANOVA results for Alzheimer’s disease–related biomarkers and vascular burden

Variables, mean (SD)	Non-APOE ε4 carriers		APOE ε4 carriers		p-value
	Without SG	With SG	Without SG	With SG
	n = 203	n = 33	n = 69	n = 15	A	G	A × G
AD-related biomarkers
Aβ composite biomarker	−0.16 (0.73)	−0.21 (0.63)	0.41 (0.71)	0.59 (0.74)	p < 0.001	0.484	0.637
p-Tau181, pg/mL	1.90 (1.04)	1.84 (0.87)	2.20 (0.89)	2.48 (1.08)	0.034	0.538	0.562
NfL, pg/mL	24.7 (14.2)	27.1 (15.8)	23.8 (9.9)	27.9 (10.9)	0.709	0.267	0.888
GFAP, pg/mL	221.0 (136.1)	203.9 (118.7)	251.0 (93.8)	365.1 (204.6)	p < 0.001	0.029	0.008
Vascular burden factors
HbA1c, %	5.7 (0.7)	5.7 (0.6)	5.6 (0.5)	6.0 (0.7)	0.427	0.201	0.018
LDL-C, mg/dL	123.6 (34.0)	115.2 (32.4)	120.8 (41.2)	135.7 (39.8)	0.050	0.319	0.043
HDL-C, mg/dL	59.3 (16.6)	56.7 (12.6)	56.8 (15.7)	55.3 (13.9)	0.935	0.777	0.541
TC, mg/dL	204.4 (39.0)	195.6 (30.7)	206.0 (32.5)	216.2 (41.3)	0.011	0.545	0.070
VRI	1.45 (0.89)	1.73 (1.13)	1.54 (0.96)	1.67 (0.90)	0.907	0.464	0.773

SG slow gait, A APOE phenotype factor, G gait performance factor, Aβ amyloid-β, p-Tau phosphorylated Tau, NfL neurofilament light, GFAP glial fibrillary acidic protein, HbA1c hemoglobin A1c, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, TC total cholesterol, VRI vascular risk index. The significant values are indicated in bold font

Fig. 1.

Boxplot illustrating the interaction between APOE ε4 and slow gait. A Post hoc analysis of two-factor ANOVA (APOE phenotype × gait performance) for GFAP. B Post hoc analysis of two-factor ANOVA for HbA1c. C Post hoc analysis of two-factor ANOVA for LDL cholesterol. SG, slow gait; GFAP, glial fibrillary acidic protein; HbA1c, hemoglobin A1c; LDL cholesterol, low-density lipoprotein cholesterol. Crosses represent the mean values

Vascular burden factors, APOE ε4, and gait

Results of two-way ANOVAs for factors related to vascular burden are also presented in Table 2. Significant interactions between APOE phenotype and gait performance were found for HbA1c and LDL cholesterol (F_{1, 312} = 5.7, p = 0.018 and F_{1, 312} = 4.1, p = 0.043, respectively). Specifically, APOE ε4 carriers with slow gait exhibited higher HbA1c levels compared to APOE ε4 carriers without slow gait, and higher LDL cholesterol levels compared to non-APOE ε4 carriers with slow gait (Fig. 1B and C). Furthermore, a significant main effect of APOE phenotype on total cholesterol was observed (F_{1, 312} = 6.6, p = 0.011), indicating that APOE ε4 carriers had significantly higher levels of total cholesterol. No significant main effects or interactions were observed for HDL cholesterol and vascular risk index.

Mediation factor for the APOE ε4 and gait relationship

Figure 2 shows results assessing mediation effects of plasma GFAP, HbA1c, and LDL cholesterol on the association between APOE ε4 carriage and slower gait speed. Elevated plasma GFAP was associated with both APOE ε4 carriage (standard coefficient [β] = 0.396, p < 0.001) and slower gait speed (β = −0.150, p < 0.001). Mediation analysis, adjusted for age, sex, ≥ 5 medication, and number of comorbidities, revealed that elevated plasma GFAP mediated the association between APOE ε4 and slower gait speed (Aroian test: p = 0.041). Specifically, the indirect effect of APOE ε4 on slower gait speed was significant due to elevated GFAP levels (effect = −0.059, 95% confidence interval = −0.12 to −0.013), while the direct effect on gait speed was diminished when plasma GFAP was included (direct effect = −0.04, p = 0.124; total effect = −0.06, p = 0.041), indicating that GFAP fully mediated this association. In contrast, neither HbA1c nor LDL cholesterol showed significant mediation effects.

Fig. 2.

Results of the mediated analysis. Elevated GFAP levels were the sole mediator of the association between APOE ε4 carriage and slower gait. When GFAP acted as a mediator, the significant direct effect was diminished, indicating full mediation. GFAP, glial fibrillary acidic protein; HbA1c, hemoglobin A1c; LDL cholesterol, low-density lipoprotein cholesterol

Underlying factor for the APOE ε4 and GFAP relationship

Additional analyses evaluating mediators of the relationship between APOE ε4 and GFAP revealed a significant indirect effect of the composite biomarker (effect = 0.23, 95% confidence interval = 0.13 to 0.37; Table 3). Specifically, inclusion of the composite biomarker diminished the direct effect of APOE ε4 on GFAP, indicating that elevated plasma Aβ levels fully mediated the association between APOE ε4 carriage and increased plasma GFAP levels. In contrast, the vascular risk index was not involved in this association (effect = 0.0001, 95% confidence interval = −0.01 to 0.02). Results remained consistent when HbA1c and LDL cholesterol were examined as mediators instead of the vascular risk index (Table S1 in the Supporting information).

Table 3.

Mediation analysis of the association between APOE ε4 and GFAP, focusing on the Aβ composite biomarker and vascular risk index

	Beta coefficients	95% CI
Each association
APOE ε4 → Aβ CB	0.60	0.43, 0.78
APOE ε4 → VRI	−0.01	−0.22, 0.20
Aβ CB → GFAP	55.75	35.25, 76.25
VRI → GFAP	−1.48	−18.66, 15.69
Total and direct effects
APOE ε4 → GFAP (total effect)	57.81	23.87, 91.74
APOE ε4 → GFAP (direct effect)	24.27	−10.56, 59.10
Partially standardized relative indirect effect(s)
APOE ε4 → Aβ CB → GFAP	0.23	0.13, 0.37
APOE ε4 → VRI → GFAP	0.0001	−0.01, 0.02

Aβ CB amyloid-β composite biomarker, VRI vascular risk index, GFAP glial fibrillary acidic protein. The significant values are indicated in bold font. Mediation factors were simultaneously included in the model. The analysis was adjusted by age, sex, medication use (≥ 5 medications), and number of comorbidities

Discussion

In this study assessing AD biomarkers in older adults with MCI participating in the J-MINT trial, a multicenter randomized controlled trial in Japan, we found that only elevated GFAP levels mediated the association between APOE ε4 carriage and slower gait. This APOE ε4–associated elevation in GFAP levels was mediated by increased plasma Aβ levels, a hallmark of AD pathology. These results suggest that APOE ε4–related gait changes may reflect AD pathology, with elevated GFAP levels, accelerated by APOE ε4, potentially contributing to reduced gait speed. The study also demonstrated that APOE ε4 carriers with slow gait showed significantly higher HbA1c and LDL cholesterol levels. However, these variables did not mediate the association between APOE ε4 carriage and slower gait, suggesting that their role is limited and reflects co-pathology rather than a causal mediator.

APOE ε4, GFAP, and gait speed

Activated glia in the form of reactive astrocytes, resulting in elevated GFAP levels, is one of the most prominent neuropathological features of AD, often found surrounding Aβ plaques in postmortem brain tissue [42]. Although GFAP levels are also known to increase with vascular risk and vascular disease [40, 43], our additional analyses confirmed that the APOE ε4–related elevation in GFAP observed in this study was primarily attributable to AD pathology. The role of these reactive astrocytes is not well understood, with some studies suggesting they are part of an endogenous defensive mechanism to eliminate the plaques, while others argue that their persistent activation induces a toxic inflammatory process that contributes to worsening AD progression [44]. Our results demonstrated that elevated GFAP, with the aforementioned characteristics, strongly explains the association between APOE ε4 carriage and slower gait among older adults with MCI. Although caution should be exercised when interpreting these results because of large variance in the value of GFAP, this finding suggests that a prodromal feature in AD development, accelerated by APOE ε4, may decrease lower extremity function, as represented by gait performance. Although this study does not specify the brain region affected by high GFAP expression in reactive astrocytes, regions involved in gait may suffer from overall inflammation. Indeed, blood GFAP levels have been also suggested as a surrogate measure of the extent of white matter injury [45, 46], which is associated with gait disturbances. Abnormal neural networks in the brain, occurring in parallel with increased reactive astrocytes, may impair gait performance.

Contrary to our study showing no association, the current body of research supports a connection between amyloid deposition and slow gait [11, 47, 48]. Considering our findings, the previously reported association between amyloid deposition and gait speed may be a secondary association that arose after the association between GFAP and gait speed was established. A significant percentage of cognitively unimpaired individuals who are Aβ-positive do not develop detectable tau pathology and, consequently, do not experience clinical deterioration [49, 50]. A recent study revealed that elevated plasma GFAP plays a role in the association of Aβ with early tau phosphorylation in preclinical AD [51]. This suggests that astrocyte reactivity can trigger the deleterious effects of Aβ in the early stages of the disease. Our results suggest that elevated GFAP levels, a factor that bridges Aβ and tau accumulation, may accelerate dementia symptoms and decrease gait speed. This assumption holds true for the previously observed association between the decline in gait speed and cognitive function. Given that gait and cognition relationships have been observed between relatively cognitively healthy older adults and those with MCI [6, 7, 52], the mediation effect of GFAP might not be specific to older adults with MCI. Furthermore, if GFAP plays a role in bridging Aβ and tau accumulation, it is reasonable that the influence of p-tau 181 and NfL on gait speed was not observed in this study, as they are considered downstream effects of Aβ and GFAP.

APOE ε4, vascular burden factors, and gait speed

We found an interaction between APOE ε4 and slow gait on HbA1c and LDL cholesterol, with higher levels of these markers among APOE ε4 carriers with slow gait. As previously rationalized for examining these, higher vascular risk burdens are well known among APOE ε4 carriers [22, 23]. Abnormal lipid and glucose metabolism has been associated with gait speed decline in the general population [53, 54]. Decreased muscle quality and strength, along with mitochondrial dysfunction, have been identified as potential causes of this relationship [55]. Considering that neither HbA1c nor LDL cholesterol mediated the association between APOE ε4 carriage and slow gait, these variables may have worsened in parallel due to APOE ε4–related changes (e.g., elevated GFAP levels), rather than directly influencing gait speed. However, it is unclear why interactive effects were observed for a limited number of variables, excluding the vascular risk index as a composite variable. The involvement of vascular burden in the association between APOE ε4 and slow gait should be examined in future studies.

Limitations

The strength of this study lies in being the first to assess the underlying pathological mechanisms of the association between APOE ε4 carriage and slow gait. However, several limitations must be acknowledged when interpreting the findings. While it is unlikely that the indirect effect observed in this study resulted from reverse causation, it remains important to examine the sequential nature of the relationship and whether the association reflects co-pathology or mediation. Potential sampling bias may have occurred because this was a secondary analysis that included only J-MINT trial participants with MCI, which could have influenced the mediating effect of AD-related pathology on the observed association. The present study assessed plasma p-tau 181 as a reliable marker of AD-tau pathology; however, this biomarker is not perfect, and reports suggest that p-tau 217 may be a more effective marker, particularly in the early stages [56, 57]. Its use might have yielded different results. Although we adjusted for potential covariates, unknown factors may still have influenced the outcomes. The analyses were adjusted for sex, given its potential influence on the pathophysiology of MCI and AD; however, a sex-stratified sub-analysis, which was not feasible due to an uneven male-to-female ratio, might have produced different findings. Moreover, the mediators examined were limited, and the possibility of unmeasured mediating variables cannot be excluded. Further longitudinal studies are warranted to confirm these findings and address these limitations.

Conclusion

The present study, a second analysis of the baseline assessment of the J-MINT trial, revealed that APOE ε4 carriers among older adults with MCI showed slower gait, consistent with prior findings. Those with APOE ε4 and slow gait had significantly higher plasma GFAP levels among AD-related biomarkers. This elevated plasma GFAP mediated the association between APOE ε4 carriage and slow gait. Given that GFAP bridges the gap between Aβ and tau accumulation, which may accelerate dementia symptoms, our results and known findings of the association between slow gait and cognitive impairment could be reasonable.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

Conceptualization: RS; methodology: RS, KU, KF, AN, TS, HA; formal analysis: RS, HS, YF; investigation: KU, KF, AN, TS, HA; writing—original draft preparation: RS, HS; writing—review and editing: YF, KU, KF, AN, TS, HA; funding acquisition: TS, HA; supervision: YF, AN, TS, HA.

Funding

This work was financially supported by the Japan Agency for Medical Research and Development (AMED) under Grant Number JP19de0107002. The sponsors had no role in the study design, data collection, analysis, and interpretation, manuscript preparation, or manuscript review and approval.

Data Availability

The data are publicly available upon request to J-MINT’s scientific publication committee at jmint@ncgg.go.jp.

Declarations

Competing interests

National Center for Geriatrics and Gerontology (NCGG), the employer of Dr. Akinori Nakamura, has filed and registered the following patents: AU 2016322342, AU 2019205010, CN 201680053338.0, EP 16846400.6, EP 19192797.9, IN 201847013746, JP 6467512, and US 15/752498. Dr. Akinori Nakamura is the inventor of these patents, but all patent rights have been transferred to NCGG. The authors report no competing interests regarding this research.