Late life physical activity relates to brain tissue synaptic integrity markers in older adults

Abstract

INTRODUCTION.

Physical activity (PA) is widely recommended for age-related brain health, yet its neurobiology is not well understood. Animal models indicate PA is synaptogenic. We examined the relationship between PA and synaptic integrity markers in older adults.

METHODS.

404 decedents from Rush Memory and Aging Project completed annual actigraphy monitoring (M visits=3.5±2.4) and postmortem evaluation. Brain tissue was analyzed for presynaptic proteins (synaptophysin, synaptotagmin-1, VAMP, syntaxin, complexin-I, and complexin-II), and neuropathology. Models examined relationships between late-life PA (averaged across visits), and timing-specific PA (time to autopsy) with synaptic proteins.

RESULTS.

Greater late-life PA associated with higher presynaptic protein levels (0.14<β<0.20), except complexin-II (β=0.08). Relationships were independent of pathology but timing-specific; participants who completed actigraphy within 2-years of brain tissue measurements showed largest PA-to-synaptic protein associations (0.32<β<0.38). Relationships between PA and presynaptic proteins were comparable across brain regions sampled.

DISCUSSION.

PA associates with synaptic integrity in a regionally-global, but time-linked nature in older adults.

Introduction

Late life physical activity (PA) is one of the most consistently recommended lifestyle modifications to support brain and cognitive aging[1]. PA is associated with a reduced incidence of Alzheimer’s dementia[2], and inactivity alone is estimated to account for >4 million dementia cases[3]. Although not synonymous with free-living activity, human exercise trials also demonstrate a positive effect on cognition[4] and gray matter growth[5], potentially supporting directionality of PA effects. Yet, the field does not fundamentally understand the biological mechanisms linking PA to brain health in humans. Identification of these pathways will not only inform risk-stratification and monitoring tools for more precise exercise recommendations, but potentially identify biological targets of cognitive resilience for future dementia prevention therapies.

Pioneering but decades-old studies have established the causal impact of PA on synaptogenesis in animals[6], an effect consistently replicated[7–10]. Voluntary PA (e.g., access to running wheel) in early, middle, or aged rodents demonstrates robust increases in dendritic tree complexity, synaptophysin[11], Ca+ signaling[12], and trophic factors (e.g., BDNF)[13]. Furthermore, the beneficial effects of PA on synaptic-related functioning generalizes across disease models (e.g., wildtype aging, amyloidosis, APOEe4, tauopathy), and corresponds with improved behavioral outcomes[9,11,14,15].

Synaptic functioning is an appealing therapeutic target given its proximity to cognition. Regardless of pathology presence, cognition cannot occur without integrity of the synaptic unit. Furthermore, preservation of synaptic structure, protein levels, and gene expression uniquely differentiates adults who are clinically resilient to AD pathology burden compared to non-resilient AD peers[16–18]. Therefore, understanding the link between PA and markers of synaptic health in humans will both begin to translate a fundamental biology observed in animal models, and build support for a readily accessible approach to support synaptic health with age.

Exercise training is associated with gray matter growth in humans[5] and the synapse is a structural component of gray matter, supporting further study of these relationships. However, the molecular balance of synaptic functioning is highly dynamic, follows non-linear patterns, and is not fully captured by brain volume metrics[19,20]. In the one human study to our knowledge, Jensen and colleagues (2017) tested the effect of a 16-week exercise intervention on cerebrospinal fluid (CSF) synaptic protein levels (neurogranin and VILIP-1) in a small cohort of adults with mild-to-moderate Alzheimer’s disease[21]; significant changes were not reached, though there was a trend for increased protein levels with high intraindividual variability. Further work is needed to carefully dissect these relationships.

We aimed to bridge this translational gap and determine the relationship between objectively measured free-living PA and synaptic markers in brain tissue of older adults. We analyzed 404 decedents from the Rush Memory and Aging Project (MAP) who completed accelerometer-based actigraphy monitoring in late life and went to autopsy with brain tissue analyzed for presynaptic proteins (synaptophysin, synaptotagmin-1, VAMP, SNAP-25, syntaxin, complexin-I, complexin-II). Building on prior works from Rush MAP that have demonstrated significant associations between actigraphy monitoring and cognition[2,22], as well as between synaptic proteins and cognition in late life[23–26], we hypothesized that late life PA would relate to higher abundances of presynaptic proteins in human brain tissue.

Methods

Participants.

404 decedents from the Rush Memory and Aging Project (MAP) were included[27]. All participants included underwent comprehensive neuropathological evaluations with available synaptic protein markers and completed actigraphy monitoring. Exclusion criteria was inability to sign an informed consent and Anatomical Gift Act.

Actigraphy Monitoring.

An activity monitor was worn on the nondominant wrist and measured rest/activity continuously (24 hours a day) for up to 10 days (Actical; Mini Mitter, Bend, OR). Activity counts were extracted for each 15-second epoch yielding 5760 data points/day. Incomplete days of data were excluded from analyses. Incomplete data were determined based on inspection of the recordings via an automated program that flagged average daily counts at the extremes: ~0/day or >500/day. Only participants with valid data for 1+ days were included in analyses. Both weekday and weekend time were captured in the 10-day monitoring period. Daily physical activity, which summarized both exercise and non-exercise activities, was calculated as the average sum of all daily activity counts for each 15-second epoch for full days of data[2].

Actigraphy monitoring was completed annually. All participants completed one visit and 73% completed 2+ visits (M visits= 3.5, SD=2.4). We examined overall late life PA levels by averaging daily activity counts across all available visits per participant, as well as visit-specific daily activity counts.

Synaptic protein quantification.

Frozen gray matter samples were obtained from six brain areas (hippocampus, middle frontal cortex, inferior temporal cortex, calcarine cortex, ventromedial caudate, and posterior putamen), and used to prepare homogenates at a consistent protein concentration, followed by serial dilution for ELISA[28]. We evaluated both a regionally global index of synaptic protein levels (aggregated across all regions sampled, for each protein), consistent with prior MAP studies,[24,28] as well as clustered, region-specific synaptic protein levels. We examined the following regions: hippocampus, cortical (middle frontal cortex and inferior temporal cortex), striatal (ventromedial caudate and posterior putamen), and calcarine cortex. Region selection was based on conceptual links with PA (i.e., striatal regions associated with movement), prior literature implicating the region with PA (i.e., hippocampus[5]), and to determine specificity against regions less commonly affected by pathology (i.e., calcarine).

Monoclonal antibodies quantified synaptophysin, synaptotagmin-1, SNAP-25, syntaxin, VAMP, complexin-I, and complexin-II levels (Supplemental Table 1). Values were expressed in log10 units, standardized, and averaged across regions within each participant. Individual protein levels provide information regarding integrity of the presynaptic compartment, higher values indicate more protein available.

Neuropathological Evaluation. Neuropathological Evaluation.

Brain removal, tissue sectioning and preservation, and uniform gross and microscopic examination followed a standard protocol previously reported[27]. All staff were blinded to clinical diagnosis. Nine common age-related pathology indicators were quantified: global burden of Alzheimer’ disease (AD) pathology, hippocampal sclerosis, Lewy Body disease (LBD), TDP-43, cerebral amyloid angiopathy (CAA), severity of arteriosclerosis and atherosclerosis, and macro- and micro-infarct counts.

Other Covariates.

See Supplemental Methods for measure details. We adjusted all models for objective motor function performances (motor 10)[2] (i.e., lifestyle PA levels accounting for motor function limitations), and conducted sensitivity analyses additionally adjusting for reported late life cognitive[29] and social activities[30] and depressive symptoms[31].

Averaged covariates.

Covariates were averaged across participant visits when included in models testing the relationship between “late life physical activity” (i.e., averaged PA across visits) and synaptic protein level. Analyses examining visit-specific actigraphy effects included visit-specific covariate values (not averaged across visits).

Statistical Analyses.

Average Late Life PA.

First, we explored bivariate associations between late life PA (averaged across visits) with demographic and clinical factors. We next evaluated the overall relationship between late life PA with global synaptic protein levels at death using multivariable regression models. All models adjusted for age at death, sex, education, and average late life motor function (motor10 averaged across visits). Next, we probed the robustness of the observed relationship between late life PA and synaptic proteins by entering additional covariates, including average late life cognitive activity, social activity, and depression levels, post-mortem interval, as well as nine common neuropathologies (i.e., AD, TDP-43, hippocampal sclerosis, LBD, macro and micro-infarcts, CAA, arteriolosclerosis, atherosclerosis). After plotting the models, a small subset of participants demonstrated disproportionately high PA levels. To confirm our results were not driven by this small subset, we conducted sensitivity analyses excluding participants with actigraphy >90^th%ile (average daily count>3.5; n=30).

Temporality.

Given putatively dynamic nature between PA and synaptic functioning in animal models showing both acute and long-term effects[32], we aimed to determine how time between measurement of in-vivo PA and post-mortem synaptic protein may affect the observed relationships. We identified participants’ last actigraphy visit and identified those who fell within ≤1 year (n=93), 1–2 years (n=119), 3–4 years (n=134), 5–6 years (n=101), or 9–10 years (n=33) to death (i.e., synaptic protein measurement; Table 3). We conducted parallel multivariable regression models examining visit-specific PA with synaptic protein levels stratified by time to death, adjusting for age at visit, education, sex, and motor function at visit. To determine whether the effects of visit were specific to the synaptic proteins versus an artifact common to other brain tissue metrics at death, we conducted the same multivariable models but instead estimated the relationship between visit-specific PA and nine common neuropathologies.

Table 3.

The relationship between in-vivo physical activity and brain tissue synaptic protein levels at death depends on time between measurements.

	Synaptophysin		Synaptotagmin1		Syntaxin		VAMP		SNAP-25		Complexin-I		Complexin-II
	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta
Average Physical Activity (n=404)	0.11 (0.04)	0.13*	0.11 (0.05)	0.13*	0.10 (0.12)	0.12*	0.12 (0.05)	0.15**	0.10 (0.04)	0.12*	0.05 (0.03)	0.07	0.02 (0.03)	0.04
Actigraphy ≤1yr to death (n=93)	0.37 (0.11)	0.36**	0.35 (0.11)	0.34**	0.33 (0.11)	0.32**	0.36 (0.11)	0.34**	0.39 (0.11)	0.38**	0.28 (0.08)	0.35**	0.03 (0.08)	0.04
Actigraphy 1–2yr to death (n=119)	0.24 (0.08)	0.31**	0.26 (0.08)	0.32**	0.19 (0.08)	0.24*	0.22 (0.08)	0.26**	0.21 (0.08)	0.27**	0.15 (0.06)	0.26*	−0.06 (0.16)	−0.1
Actigraphy 3–4yr to death (n=134)	0.07 (0.06)	0.11	0.06 (0.06)	0.09	0.04 (0.06)	0.06	0.07 (0.07)	0.10	0.05 (0.06)	0.07	0.02 (0.06)	0.03	0.01 (0.05)	0.03
Actigraphy 5–6yr to death (n=101)	−0.003 (0.05)	−0.005	−0.01 (0.05)	−0.02	0.01 (0.05)	0.02	−0.01 (0.05)	−0.02	−0.46 (0.31)	−0.09	−0.01 (0.05)	−0.16	−0.09 (0.12)	−0.15
Actigraphy 9–10yr to death (n=33)	−0.29 (0.17)	−0.37	−0.24 (0.18)	−0.31	−0.12 (0.18)	−0.16	−0.22 (0.17)	−0.28	−0.14 (0.16)	−0.2	−0.18 (0.13)	−0.31	−0.17 (0.11)	−0.31

Note.

^*p<0.05;

^**p<0.01;

each column represents a regression model. Average Late Life Actigraphy models adjusted for age at death, sex, education, and average late life motor10 performance. All other models adjusted for age at visit, sex, education and motor10 performance at visit.

To help determine if effects were related to pathology accumulation (e.g., reverse causality), we tested the relationship between PA and synaptic proteins adjusting for the interactions between PA with clinical severity at death (normal vs. MCI vs. dementia) or with individual neuropathologies on synaptic proteins.

Regionality.

Lastly, we evaluated the relationship between average late life PA and regionally-specific synaptic protein levels using parallel multivariable regression models, adjusting for age at death, sex, education, and motor function. Four a priori regions were selected (hippocampus, cortical, striatal, and calcarine).

A priori hypotheses used p-value < 0.05 to indicate statistical significance. Standardized and unstandardized beta coefficients with standard error are reported.

Results

Participants completed the actigraphy monitoring for an average of 3.5 visits (range=1–10), and averaged 90-years-old at death (Table 1). At autopsy, the cohort was roughly evenly split across clinical neurocognitive diagnoses (i.e., normal, MCI, or dementia), and 65% meet Reagan criteria for neuropathological AD.

Table 1.

Participant clinical and pathological characteristics (N=404).

ACTIGRAPHY VISITS (N, %)
1	404 (100%)
2	298 (73.8%)
3	224 (55.4%)
4	164 (40.6%)
5	111 27.5%)
6	67 (16.5%)
7	49 (12.1%)
8	27 (6.7%)
9	9 (2.2%)
10	4 (1.0%)
	mean = 3.5 (SD=2.4)
AGE AT DEATH (YEARS; M, SD)	90.6 (5.8)
AGE AT LAST ACTIGRAPHY VISIT (YEARS; M, SD)	88.5 (6.1)
SEX (% FEMALE, N)	72.1% (280)
EDUCATION (YEARS; M, SD)	14.6 (2.8)
COGNITIVE DIAGNOSIS AT DEATH (N, %)
CLINICALLY NORMAL	36.3% (141)
MCI-AD ☨	24.2% (94)
MCI-OTHER	1.8% (7)
DEMENTIA-AD ☨	33.2% (129)
DEMENTIA-OTHER	4.4% (17)
AVERAGE LATE LIFE LIFESTYLE FACTORS
ACTIGRAPHY: AVERAGE DAILY (COUNTS; M, SD)	1.97 (1.1)
MOTOR FUNCTION COMPOSITE (Z-SCORE; M, SD)	0.87 (0.17)
DEPRESSIVE SYMPTOMS (CES-D; M, SD)	1.4 (1.3)
COGNITIVE ACTIVITY (POSSIBLE RANGE 1 TO 5; M, SD)	2.98 (0.63)
SOCIAL ACTIVITY (POSSIBLE RANGE 1 TO 5; M, SD)	2.34 (0.46)
NEUROPATHOLOGICAL EXAMINATION
POSTMORTEM INTERVAL (HOURS; M, SD)	8.2 (5.1)
ALZHEIMER’S DISEASE PATHOLOGY (%, REAGAN CRITERIA, N)	64.6% (250)
BRAIN TISSUE SYNAPTIC PROTEINS (GLOBAL Z-SCORE)☨☨
COMPLEXIN-I (M, SD)	−0.20 (0.74)
COMPLEXIN-II (M, SD)	−0.17 (0.67)
SYNAPTOPHYSIN (M, SD)	−0.19 (0.92)
SNAP-25 (M, SD)	−0.18 (0.92)
SYNTAXIN (M, SD)	−0.18 (0.93)
VAMP (M, SD)	−0.18 (0.95)
SYNAPTOTAGMIN-1 (M, SD)	0.02 (0.95)

^☨NINCDS consensus criteria for possible or probable AD

^☨☨Brain synaptic protein values are expressed in log10 units, standardized to the MAP sample, and averaged across six regions within each participant (z-score units reported).

Note. All lifestyle factors represent an average of participants values across visits to capture illustrate robust indicators of late life behaviors. MCI = Mild Cognitive Impairment; AD = Alzheimer’s disease; CES-D = Center for Epidemiologic Studies – Depression scale. M = mean; SD = standard deviation.

Does average late life physical activity relate to synaptic protein levels at death?

On average, participants wore the Actical device for 9.3 days (SD= 1.3) over the annual 10-day monitoring periods. Length of monitoring did not differ by proximity to death (M= 9.4 days for those ≤1 year to death and M= 9.2 days for those 7+ years to death). Average late life PA demonstrated small-to-minimal associations with age at death (r=−0.13, p=0.008), education (r=−0.06, p=0.17), and sex (t(402)=0.43, p=0.65). Greater PA related to better late life motor function performances (r=0.39, p<0.001).

Adjusting for demographics and average motor function, greater late life PA was associated with higher synaptophysin (B=0.13, p=0.017), VAMP (B=0.11, p<0.001), SNAP-25 (B=0.12, p=0.026), synaptotagmin-1 (B=0.13, p=0.014), and syntaxin (B=0.12, p=0.029) levels at death (Figure 1). Average PA did not significantly relate to complexins (complexin-I B=0.07, p=0.18; complexin-II B=0.04, p=0.45). No interactions between sex and PA were found (p-values>0.05).

Figure 1.

Greater average late life physical activity relates to higher synaptic protein in brain tissue at death, with the exception of complexin-II.

Note. Dashed lines represent 95%CI. Average late life physical activity represents actigraphy levels averaged across available study visits. All models adjusted for age at death, sex, education, and average late life motor10 performances. Synaptic protein levels are expressed in log10 units, standardized, and averaged across six brain regions within each participant (z-score units).

To test specificity of PA compared to other modifiable factors, we adjusted models for average late life social activity, cognitive activity, and depression symptomology. Only daily PA continued to significantly relate to the same synaptic proteins to a similar degree (Table 2). Given published associations between synaptic integrity and pathology burden[28], we further adjusted for nine common neuropathologies and post-mortem interval. The PA model coefficients became slightly stronger and now greater late life PA additionally significantly related to higher complexin-I, but not complexin-II levels (Table 2).

Table 2.

Average late life physical activity is independently associated with brain tissue synaptic protein levels at death (n=404).

	Estimate	Synaptophysin	Synaptotagmin-1	Syntaxin	VAMP	SNAP-25	Complexin-I	Complexin-II
Avg Physical Activity	Coeff (se)	0.15 (0.04)	0.16 (0.05)	0.14 (0.04)	0.17 (0.05)	0.14 (0.04)	0.09 (0.03)	0.05 (0.03)
	Std Beta	0.18**	0.19**	0.17**	0.20**	0.17**	0.14*	0.08
Age at death	Coeff (se)	−0.01 (0.01)	−0.01 (0.01)	−0.004 (0.01)	0.002 (0.01)	−0.003 (0.01)	−0.02 (0.01)	−0.01 (0.01)
	Std Beta	−0.05	−0.04	−0.03	0.01	−0.02	−0.13*	−0.09
Sex	Coeff (se)	0.09 (0.11)	0.10 (0.11)	0.90 (0.11)	0.04 (0.11)	0.13 (0.11)	−0.03 (0.09)	−0.17 (0.08)
	Std Beta	0.003	0.05	0.05	0.02	0.06	−0.02	−0.11
Education	Coeff (se)	0.001 (0.02)	−0.01 (0.02)	−0.01 (0.02)	−0.001 (0.02)	−0.01 (0.02)	0.0001 (0.01)	0.002 (0.01)
	Std Beta	0.003	−0.02	−0.02	−0.002	−0.03	0.003	0.001
Avg Motor10	Coeff (se)	−0.44 (0.31)	−0.34 (0.31)	−0.38 (0.31)	−0.25 (0.32)	−0.46 (0.31)	−0.02 (0.24)	−0.11 (0.24)
	Std Beta	−0.08	−0.06	−0.07	−0.05	−0.09	−0.005	−0.03
Avg Cognitive Activity	Coeff (se)	0.02 (0.08)	0.04 (0.09)	0.01 (0.09)	0.03 (0.09)	0.03 (0.08)	0.02 (0.07)	0.08 (0.06)
	Std Beta	0.01	0.03	0.01	0.02	0.02	0.02	0.08
Avg Social Activity	Coeff (se)	−0.07 (0.12)	−0.11 (0.12)	−0.11 (0.12)	−0.11 (0.12)	−0.09 (0.12)	−0.25 (0.09)	−0.20 (0.09)
	Std Beta	−0.04	−0.05	−0.05	−0.05	−0.04	−0.16**	−0.14*
Avg CES-D	Coeff (se)	−0.04 (0.04)	−0.04 (0.04)	−0.06 (0.04)	−0.06 (0.04)	−0.04 (0.04)	0.03 (0.03)	−0.01 (0.03)
	Std Beta	−0.06	−0.06	−0.08	−0.08	−0.07	−0.05	−0.02
Global AD	Coeff (se)	−0.19 (0.08)	−0.17 (0.09)	−0.20 (0.08)	−0.17 (0.08)	−0.16 (0.08)	−0.24 (0.07)	−0.14 (0.06)
	Std Beta	−0.12*	−0.11*	−0.13*	−0.11*	−0.10	−0.19**	−0.12*
Hippocampal sclerosis	Coeff (se)	0.25 (0.17)	0.31 (0.18)	0.22 (0.18)	0.23 (0.18)	0.20 (0.17)	−0.02 (0.14)	−0.31 (0.13)
	Std Beta	0.08	0.09	0.07	0.07	0.06	−0.007	−0.13*
DLB	Coeff (se)	−0.02 (0.04)	−0.01 (0.04)	−0.02 (0.04)	−0.04 (0.04)	−0.02 (0.04)	−0.02 (0.03)	−0.01 (0.03)
	Std Beta	−0.03	−0.01	−0.03	−0.04	−0.02	−0.03	−0.02
TDP-43	Coeff (se)	0.02 (0.04)	0.004 (0.04)	0.04 (0.04)	0.05 (0.04)	0.03 (0.04)	−0.02 (0.03)	0.06 (0.03)
	Std Beta	0.02	0.01	0.05	0.06	0.04	−0.03	0.11*
Arteriolosclerosis	Coeff (se)	0.18 (0.06)	0.17 (0.06)	0.14 (0.05)	0.17 (0.06)	0.16 (0.06)	0.04 (0.04)	0.06 (0.04)
	Std Beta	0.17**	0.16**	0.13*	0.15*	0.15**	0.05	0.08
CAA	Coeff (se)	0.04 (0.05)	0.01 (0.05)	0.03 (0.05)	0.03 (0.05)	0.04 (0.05)	0.03 (0.04)	0.03 (0.04)
	Std Beta	0.04	0.01	0.03	0.03	0.04	0.04	0.04
CVDA	Coeff (se)	0.24 (0.06)	0.26 (0.06)	0.23 (0.06)	0.20 (0.06)	0.20 (0.06)	0.24 (0.05)	0.05 (0.04)
	Std Beta	0.22**	0.23**	0.20**	0.17**	0.18**	0.27**	0.06
Macroinfarcts	Coeff (se)	−0.01 (0.10)	0.02 (0.10)	−0.001 (0.10)	0.05 (0.10)	0.06 (0.10)	−0.11 (0.08)	−0.06 (0.07)
	Std Beta	0.003	0.12	−0.0004	0.02	0.03	−0.07	−0.05
Microinfarcts	Coeff (se)	−0.14 (0.10)	−0.14 (0.10)	−0.23 (0.10)	−0.15 (0.10)	−0.17 (0.10)	−0.14 (0.08)	−0.03 (0.08)
	Std Beta	−0.07	−0.07	−0.12*	−0.08	−0.09	−0.09	−0.02

Note.

^*p<0.05;

^**p<0.01;

columns represent individual regression models. Avg = averaged across study visits. AD = Alzheimer’s disease; CES-D = Center for Epidemiologic Studies – Depression scale; DLB = dementia with Lewy Body pathology staging; TDP-43 = TAR DNA-binding protein-43; CAA = cerebral amyloid angiopathy; CVDA = cerebral atherosclerosis.

To test veracity of the models excluding possible outliers, we re-ran models excluding participants with >90^th%ile of daily movement. Model estimates remained similar (B range=0.13 to 0.17; p<0.03).

Temporality: Does time between physical activity and synaptic measurement affect the relationship?

Given animal models suggesting the relationship between PA and synaptic outcomes may show both acute and longer-term effects, we stratified the cohort by years to death to better understand the temporal dynamics of this relationship in humans. Adjusting for demographics and motor function at visit, the relationship between PA and synaptic protein levels demonstrated a dose-dependent increase in size as participants neared the time of synaptic protein measurement (i.e., death). There was a positive, medium-sized relationship between PA and all synaptic protein markers, with the exception of complexin-II, in participants 0–2 years from death that attenuated the further out the actigraphy measurement was taken. Together, these data suggest that proximity between actigraphy and synaptic protein level measurement may be an important factor (Table 3; Figure 2).

Figure 2.

The relationship between physical activity and brain tissue synaptic protein levels is stronger the more proximate the measures were taken in time (i.e., time between actigraphy monitoring and death).

Note. N=150 participants 0–2 years to death and n=154 participants 3–10 years to death. Fitted regression lines adjusting for age at visit, sex, education, and motor10 performance at visit are plotted with 95% CI.

To test if this was an artifact common to other brain tissue metrics at death, we evaluated the relationship between last PA measurement and neuropathologies, similarly stratified by years to death. We did not observe a consistent change in the size of relationship between PA and neuropathology as a function of time to death (Supplementary Table 2), suggesting our visit-related findings were more specific to synaptic protein levels.

We further probed if the stronger relationship between PA and synaptic protein in participants 0–2 years from death reflected effects of pathology. In participants 0–2 years from death (n=150), the size of the relationship between PA and synaptic protein levels did not substantially change when adjusting for nine common neuropathologies (Synaptophysin B=0.29; Synaptotagmin-1 B=0.29; Syntaxin B=0.24; VAMP B=0.25; SNAP-25 B=0.28; Complexin-I B=0.25; Complexin-II B=0.08), and there were no significant interactions between PA and individual pathologies (e.g., not driven by individuals with AD pathology; all p-values >0.05). There was, however, a significant interaction between PA and clinical diagnosis at death on levels of synatotagmin-1, syntaxin, SNAP-25, and complexin-I (interaction term p-values <0.04). Stratified analyses showed positive associations between PA and these synaptic proteins in all three diagnosis, though the effect size was strongest in participants diagnosed with MCI (Supplemental Figure 1).

Regional specificity: Does the relationship between physical activity and synaptic proteins differ across the brain?

Adjusting for demographics and average motor function, greater average late life PA related to higher synaptic protein levels relatively equally across brain regions (Table 4). Given the important time of measurement effects observed, we also evaluated these relationships stratified by participants 0–2 years to death (n=150) compared to those 3+ years from death (n=154). Again, we found greater PA related to higher synaptic protein levels most strongly in those 0–2 years to death, an effect that appeared similar across brain regions for all synaptic proteins (Table 4). In other words, no brain region appeared to be driving or susceptible to the observed timing effects.

Table 4.

Models demonstrating the size of the relationship between physical activity and brain tissue synaptic protein levels appears similar across brain regions.

	Synaptophysin		Synaptotagmin1		Syntaxin		VAMP		SNAP-25		Complexin-I		Complexin-II
	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta	Coeff (se)	Std Beta
Average Late Life Actigraphy (all participants, n=404)
Striatal	0.14 90.05)	0.16**	0.15 (0.05)	0.17**	0.10 (0.05)	0.12*	0.16 (0.05)	0.18**	0.12 (0.05)	0.14*	0.05 (0.04)	0.06	0.02 (0.04)	0.03
Hippo	0.10 (0.05)	0.12*	0.11 (0.05)	0.13*	0.10 (0.05)	0.11*	0.11 (0.05)	0.12*	0.09 (0.05)	0.11	0.08 (0.05)	0.09	0.01 (0.05)	0.01
Cortical	0.10 (0.05)	0.12*	0.10 (0.05)	0.11*	0.09 (0.05)	0.10	0.11 (0.05)	0.13*	0.10 (0.05)	0.12*	0.06 (0.04)	0.07	0.02 (0.04)	0.03
Occipital	0.12 (0.05)	0.14*	0.13 (0.05)	0.15**	0.11 (0.05)	0.13*	0.13 (0.05)	0.15**	0.12 (0.05)	0.14*	0.05 (0.05)	0.05	0.07 (0.05)	0.08
Actigraphy 0–2yr to death (n=150)
Striatal	0.27 (0.07)	0.33**	0.27 (0.07)	0.32**	0.22 (0.07)	0.28**	0.26 (0.08)	0.30**	0.22 (0.07)	0.27**	0.12 (0.07)	0.16	−0.07 (0.06)	−0.10
Hippo	0.29 (0.07)	0.34**	0.32 (0.08)	0.37**	0.26 (0.07)	0.31**	0.25 (0.08)	0.31**	0.29 (0.08)	0.34**	0.30 (0.08)	0.33**	−0.01 (0.07)	−0.01
Cortical	0.26 (0.07)	0.31**	0.26 (0.08)	0.31**	0.22 (0.07)	0.27**	0.23 (0.08)	0.27**	0.25 (0.07)	0.30**	0.18 (0.07)	0.23**	−0.03 (0.07)	−0.04
Occipital	0.28 (0.07)	0.33**	0.29 (0.08)	0.33**	0.24 (0.08)	0.30**	0.27 (0.07)	0.32**	0.28 (0.08)	0.32**	0.17 (0.08)	0.20*	0.01 (0.08)	0.01
Actigraphy 3+yr to death (n=154)
Striatal	0.17 (0.08)	0.21*	0.20 (0.08)	0.25**	0.13 (0.07)	0.17	0.23 (0.07)	0.30**	0.15 (0.08)	0.19	0.13 (0.07)	0.16	0.02 (0.07)	0.03
Hippo	0.13 (0.08)	0.15	0.21 (0.08)	0.26**	0.14 (0.08)	0.17	0.17 (0.07)	0.22*	0.17 (0.08)	0.20*	0.08 (0.09)	0.09	0.001 (0.08)	0.001
Cortical	0.14 (0.08)	0.17	0.17 (0.08)	0.21*	0.11 (0.08)	0.13	0.19 (0.08)	0.23*	0.15 (0.08)	0.18	0.11 (0.07)	0.15	0.02 (0.07)	0.03
Occipital	0.09 (0.08)	0.10	0.17 (0.08	0.20*	0.10 (0.08)	0.12	0.16 (0.08)	0.18	0.09 (0.08)	0.11	0.11 (0.08)	0.14	0.08 (0.08)	0.10

Note.

^*p<0.05;

^**p<0.01;

Average Late Life Actigraphy models adjusted for age at death, sex, education, and average late life motor10 performance. All other models adjusted for age at visit, sex, education and motor10 performance at visit. “Cortical” represents middle frontal and inferior temporal gyri. Hippo = hippocampal.

Discussion

Physical activity (PA) is consistently linked to brain health, yet the biological pathways supporting this relationship in humans are not well understood. We demonstrate that greater free-living PA levels, measured by objective actigraphy monitoring in life, associate with presynaptic protein levels of synaptophysin, synaptotagmin-1, syntaxin, VAMP, SNAP-25, and complexin-I in brain tissue at death. Although this relationship was initially small, it was robust and statistically independent from overall motor function (i.e., ability to engage in PA), other modifiable factors (i.e., cognitive and social activities, depression), and neuropathology burden. Interestingly, the association between PA and presynaptic proteins became strikingly larger the closer the measurements were taken in time. The strongest effect sizes were observed in participants 0–2 years to death, a relationship disproportionately evident among individuals diagnosed with MCI before death. Additionally, this appeared to be a regionally global phenomenon with PA relating to synaptic protein levels relatively equally across brain regions sampled. Despite adjusting for confounding factors, due to the observational study design, we cannot fully determine directionality of effects; it is likely the relationship between PA and synaptic integrity markers are bidirectional with at least some contribution from reverse causality (i.e., declines in synaptic functioning leading to lower PA engagement). Nonetheless, these are the first data to demonstrate a relationship between PA and markers of synaptic integrity in human brain tissue and suggest: 1) PA may support and/or build synaptic health in a temporally dynamic nature, 2) PA interventions may have optimal therapeutic windows for synaptic outcomes and be particularly effective in individuals in transitional cognitive states (MCI), and 3) PA may be beneficial across brain networks supporting cross-diagnostic utility.

Synapses are the biologic correlate of cognition and therefore an appealing therapeutic target. Our data show that greater PA relates to higher levels of presynaptic proteins, suggesting that PA may maintain or build brain resilience (i.e., more synaptogenesis, and/or increased synaptic protein reservoir) through processes involved in vesicle formation and trafficking, and regulation of neurotransmitter release. We observed some specificity, such that complexin-II did not demonstrate meaningful relationships with PA. Complexin-II is primarily present in glutamatergic terminals and prior analyses from our group suggest it is more affected and most strongly associated with cognition in later stages of disease[26]. This is in contrast to the association we did observe between PA and complexin-I, which more highly represented in inhibitory terminals, and more strongly relates to cognition, particularly in earlier disease states[26]. Interestingly, animal models have shown increases in excitatory (glutamatergic) and decreases in inhibitory (GABA) transcriptional pathways acutely following exercise[7]. Our findings indicating specific relevance of inhibitory terminals may reflect differences in the timing (acute vs. years), age, and/or species examined. Our data are also closely aligned with extant animal experiments demonstrating the causal benefit of exercise on synaptogenesis, as well as induction of protein or mRNA levels of the specific presynaptic proteins we examined (synaptophysin, SNAP-25, synaptotagmins, and syntaxin)[33–38]. For example, using a data-driven approach (microarray reflecting >1000 cDNAs), Molteni and colleagues (2002) demonstrated the largest changes following PA were observed for upregulation in presynaptic trafficking pathways, including synaptotagmin and syntaxin. Interestingly, blockade of TrkB (BDNF receptor) inhibits exercise-induced increases in cognition, as well as synaptic protein levels (e.g., synaptotagmin, synaptophysin, synapsin[36–38]), highlighting a critical role for BDNF coordinating PA-induced synaptic changes. Given prior exercise animal models have also implicated molecules in the postsynaptic compartment (e.g., PSD95, neurogranin)[35,39], a more comprehensive understanding of the other trans- and post-synaptic pathways that may mediate PA effects on the brain in humans is needed. Careful dissection of these molecular relationships will both improve our fundamental understanding of behavior-to-brain relationships and aid in identifying possible resilience-related pathways promotable via PA or other therapies.

Other outstanding questions regarding the effects of PA on the brain center around potential durability and regionality of this relationship. We found there was a dose-dependent increase in the size of the relationship between PA and synaptic proteins the more proximal the measurements were taken in time. Medium effect sizes (standardized betas ~0.30) were observed for participants within 0–2 years to death that was not driven by neuropathology burden but was most evident for individuals diagnosed with MCI. Therefore, the relationship between PA and synaptic regulation appeared relatively tightly coupled in time, potentially indicating that initiation and/or continuity of higher activity levels may be needed to support ongoing synaptic health. Additionally, individuals who are just beginning to display cognitive symptoms (MCI) may represent a particularly plastic therapeutic window of opportunity ideal for PA interventions targeting synaptic functioning. Consistent with this latter finding, prior works have demonstrated an initial “compensatory” increase in synaptic protein levels among individuals and animals in the earliest stages of neurodegeneration[40,41]. Again, it is important to note that given our observational design, it is not possible to determine if PA is driving synaptic health or vice versa. Nonetheless, these data help set the stage for the human intervention work that is needed to untangle the true temporal dynamics and person-specific predictors (e.g., disease stage) of those who stand to benefit most.

Although many prior animal (e.g.,[9,34,35]) and human exercise studies (e.g., [5]) have focused on the hippocampus, we show broadly comparable relationships between PA and synaptic protein levels across brain regions sampled. Perhaps this is not surprising given the systemic nature of PA. For example, at least some of the neurotrophic pathways of PA are likely through maintenance of cerebrovascular and glymphatic systems, as well as immune homeostasis, all of which are globally present throughout the brain[42]. The global nature of the PA-synaptic relationship observed is also consistent with data-driven voxel-based morphometry analyses and analyses examining whole-brain metrics that have demonstrated globally distributed relationships between PA and brain structural integrity[43,44]. Additionally, the lack of regional specificity suggests that PA may be a promising intervention across neurodegenerative syndromes (i.e., regardless of brain network affected), which is supported by an increasing array of PA studies across diagnostically diverse populations (e.g., aging, AD, FTLD, HIV, Parkinson’s disease).[5,45,46]

Our study is not without important limitations. As noted, PA and synapse relationships are likely highly dynamic and bidirectional in nature and our observational study design precludes conclusions regarding causality. This may be particularly relevant in our analyses showing temporal dependence of the PA-synapse relationship in which phenomena surrounding death may be a confounding factor (e.g., both PA and synaptic integrity may decrease closer to death). We demonstrate that PA relates to synaptic outcomes independent of chronological age at death and time to death (data not shown), and that PA did not show time-linked associations with other brain tissue outcomes and synaptic markers did not show time-linked associations with other in-vivo outcomes (data not shown). Together, this suggests specificity of temporal patterns to the PA-synaptic relationship. However, it is not possible to fully rule out effects of death on our results given our observational study design. Additionally, we focused on markers of synaptic functioning; there are likely many, diverse up- and down-stream mechanisms at play linking PA to the synapse and brain health, more broadly. Integration of molecular markers reflecting glial, immune, and endothelial functioning (among others) may be highly fruitful to understand the neurobiologic cascade of PA. Relatedly, our synaptic protein measurements are quantified in brain tissue supporting their specificity in CNS function, yet they are limited in truly capturing the time-linked dynamics of PA. Given the increasing availability and validity supporting measurement of synaptic protein levels via CSF or PET in living humans[19,47], in-vivo replication of these relationships will be an important future direction to help refine our understanding of the relevance and temporal dynamics between PA and synaptic markers. Additionally, our metric of physical activity was an aggregated composite of daily movement. We cannot determine which specific activities (e.g., structured vs. unstructured exercise, intensity) or patterns of activity may be most relevant in the observed relationships. Future studies further dissecting the characteristics of these relationships to determine the pattern(s) of physical activity most related to brain health are needed to inform more precise clinical recommendations. Lastly, it is important to note potential sample bias inherent in intensive clinicopathologic studies such as Rush MAP. Although this cohort is invaluable for discovery of potential targets and pathways of interest as measured directly in brain tissue, it may not inherently represent the broader older adult population and future works are needed to determine the generalizability of our findings.

In sum, our data are the first to demonstrate a link between a lifestyle behavior, PA, and markers of synaptic integrity in human brain tissue. We suggest PA may help build synaptic health, even at late ages (e.g., “brain reserve”)[48], independent from pathology, but that this is a potentially plastic process that may need to be sustained over time. Given its proximity to cognitive functioning and posited role in sustaining cognitive resilience to neuropathology[16–18], modifiable approaches to support synaptic health are of high relevance in our aging population. These data suggest that synaptic metrics may be particularly compelling and relevant outcomes for future behavioral intervention studies.

Research in Context.

1. Systemic review: We reviewed current literature using traditional search engines (e.g., PubMed, GoogleScholar). Physical activity (PA) is consistently linked with age-related brain structural and functional health; yet, its biological pathways are not well understood in humans. Decades of animal experiments demonstrate PA is synaptogenic. The fundamental relationship between PA and synaptic integrity has not been elucidated in humans.

2. Interpretation: Our data are the first to show PA relates to markers of synaptic integrity in human brain tissue. The relationship appeared regionally global and independent of pathology, but timing-dependent; the closer measurements were taken in time, the larger the relationship. PA may help build synaptic health, and this relationship may be a temporally dynamic process.

3. Future directions: Future works are needed to parse out in-vivo temporal dynamics of PA with synaptic markers, including intervention studies to support directionality.

Highlights.

• Do free-living physical activity levels relate to synaptic integrity in humans?

• Greater late life activity related to higher presynaptic protein levels in brain tissue.

• Relationships were strongest the more proximate the measurements were taken in time.

• Relationships were independent of pathology and comparable across brain regions.

• Physical activity may be a modifiable tool to support synaptic health.