Of Aging Mice and Men: Gait Speed Decline Is a Translatable Trait, With Species-Specific Underlying Properties

Abstract

In the last two decades, great strides were made in our ability to extend the life span of model organisms through dietary and other manipulations. Survival curves provide evidence of altered aging processes but are uninformative on what lead to that increase in life span. Longitudinal assessments of health and function during intervention studies could help in the identification of predictive biomarkers for health and survival. Comparable biomarkers of healthspan are necessary to effectively translate interventions into human clinical trials. Gait speed is a well-established predictive biomarker of healthspan in humans for risk of disability, health outcomes and mortality, and is relatively simple to assess noninvasively in rodents. In this study, we assessed and compared gait speed in males from two species (mice and humans), from young adulthood to advanced old age. Although gait speed decreases nonlinearly with age in both species, the underlying drivers of this change in gait speed were different, with humans exhibiting a shortened step length, and mice displaying a decrease in cadence. Future longitudinal and interventional studies in mice should examine the predictive value of longitudinal declines in gait speed for health and survival.

Geroscience studies the relationships between aging and age-related diseases, and addresses issues of longevity and healthspan across species. To facilitate effective translation of animal models of aging to human clinical trials, comparable biomarkers of healthspan should be used across species (1,2). Because gait speed has been shown to be a powerful predictor of many aspects of health, it has been proposed as a key biomarker of healthspan in humans (3) and across species (1). In humans, gait speed is strongly associated with mortality (4,5), current and future functional status (6–9), adverse health outcomes (10–12), and health care costs (13). In the context of health care policy and the ultimate impact of aging research on human well-being, the decline in gait speed represents a powerful contributor to the functional, medical, and health economic burden of aging worldwide. While mice are frequently used in the study of aging, age-related changes in gait speed in mice have only recently been assessed (14–17). To the extent that mouse models of aging inform pathways to intervention in the elderly adults, it is necessary to understand how aging-related changes in gait in mice resemble and/or differ from changes in humans. In this study, our aim is to explore how key gait parameters contribute to the age-related decline in gait speed across species, to form a foundation for how results of research in mice on healthspan factors reflect and differ from the human experience, and ultimately to use these insights to inform the design of preclinical interventions. Here, we assessed changes in gait in male B6 mice and men from young adulthood to advanced old age, and evaluated how distinct cross-species changes in the two main components of gait, step length and cadence, account for age-related decline in gait speed.

Methods

Human Gait Analysis

The human gait data are from the Baltimore Longitudinal Study of Aging, a study of normative human aging established in 1958, conducted by the Intramural Research Program of the National Institute on Aging (NIA) (18). The study is IRB approved and all participants provided informed consent. Participants were excluded from the analysis if they needed walking assistance, were legally blind, had cognitive impairment, a body mass index greater than 40, or a medical diagnosis known to affect gait (eg, stroke or Parkinson’s disease). The cross-sectional sample was limited to men (N = 242, age 32–89 years) because the mouse cohort was all male. To make cross species comparisons across the adult life span, we scaled our measure of age; human age is scaled to an optimal life span of 100 years. Therefore, the relative age range for humans is from 32% to 89% of the maximal life expectancy. For the gait assessment, participants were instructed to walk at their self-selected usual walking speed, looking ahead and not talking during walking, along a 10-m walkway. Body segment movements were recorded by a 10-camera Vicon motion analysis system (Oxford Metrics Ltd., UK). Step length is the distance between two consecutive heel strikes identified by heel markers, and gait frequency (ie, cadence) is the step count per minute. As a standard gait analysis approach, the first two and the last two steps in each walking trial were excluded to eliminate acceleration and deceleration phases of gait.

Mouse Gait Analysis

The mouse gait data were collected from B6 male mice obtained from the Jackson Laboratory (Bar Harbor, ME) or NIA Aging Colony and aged in-house at the Biomedical Research Center (Baltimore, MD) under approved animal study protocols. This study includes a cross-sectional cohort of male mice (N = 99, age 12.0–33.3 months). To compare with humans, mouse age is scaled to an optimal life span of 37 months, which represents the typical life span for a B6 male mouse on standard diet (19). Therefore, the relative age range for mice is from 32% to 89% of the maximal life expectancy. For the gait assessment, mice were acclimated to a corridor (120 × 5 cm) for 1 week before gait recording. A maximum of five runs per day were allowed for each mouse during training to minimize stress. During gait assessment, the observer was present, the corridor was lit, and the home cage was placed at the opposite end to motivate the mouse to walk. Two trials were conducted to ensure the mouse walked the corridor length, and one valid trial (without stopping or turning in the opposite direction in the 70 cm mid-runway filming area) was used for analysis. Paws were tracked by semiautomated computer vision software package provided by the company (TSE Systems). A custom in-house Python script was used to calculate step length and cadence from the recorded spatial coordinates and timing of the paw movements. Step length was measured as the distance one limb takes during swing phase. All four limbs were used and then averaged for the step length calculations. Cadence was calculated as number of steps per minute.

Nondimensional Normalization for Gait Parameters

In order to examine the effect of age on gait across animal species of different body size, gait parameters must be scaled (1). There is no single universal weight-based scaling equation across bipedal and quadrupedal gaits. We use an alternative approach of nondimensional normalization based on length (L¹) and time (T¹) (20). This approach has been used to study sex-specific age-related gait decline in humans to adjust different body size between men and women (21). In the context of gait, length can be normalized by body height or effective leg length (l₀, the distance from hip joint to ground). Time can be normalized using the inverse of the pendulum period (ie, $1/\sqrt{g/l_0}$, g is the acceleration due to gravity, 9.8 m/s²) to scale frequency (eg, cadence in the context of gait) because both bipedal and quadrupedal gait kinematics can be described by movement of an inverted pendulum (22,23). In this study, we used body height (for humans) or body length (for mice) as a proxy for l₀. Body length for mice is measured from neck to rump, not including the tail. Scaling for velocity is simply L¹/T¹ (ie, $1/\sqrt{g{l}_0}$). Table 1 lists the average original and scaled gait data by species. The nondimensionally normalized gait parameters are used in Figure 1 and in statistical analyses.

Table 1.

Gait Parameters in Humans and Mice.

	Gait Speed	Step Length	Cadence
Humans	119.2 ± 18.0 cm/s	64.5 ± 7.6 cm	111.3 ± 8.3 steps/min
	(0.29 ± 0.04)	(0.37 ± 0.04)	(47.06 ± 3.48)
Mice	24.6 ± 10.0 cm/s	4.6 ± 1.1 cm	500.1 ± 196.5 steps/min
	(0.29 ± 0.11)	(0.67 ± 0.13)	(42.64 ± 17.81)

Note: Original and nondimensionally normalized gait parameters (in parenthesis, unitless). Mean ± SD.

Figure 1.

LOWESS fit curves for nondimensionally normalized gait parameters in humans and mice. Note: All gait parameters are nondimensionally normalized and therefore are unitless. Age at x-axis is expressed as % of maximal life expectancy (% MLE, humans: 100 years, mice: 37 months). Left panels (A, C, E) are for human gait data; and right panels (B, D, F) are for mice gait data. Upper panels (A, B) are for gait speed; middle panels (C, D) are for step length; and lower panels (E, F) are for cadence.

Statistical Analysis

Each nondimensionally normalized gait parameter (gait speed, step length, and cadence) was analyzed by PROC GLM model with SAS 9.4. Independent factors include species, age², and a “species × age²” interaction. We chose age² rather than age as the independent factor due to the nonlinear relationship between gait parameters and age observed in the LOWESS fit curves (Figure 1). Because some gait parameters show a greater decline at older age, we further examined the age effect by age group (Younger: <65; and Older: >65 maximal life expectancy) for each species.

Results

Table 1 lists the original and nondimensionally normalized gait parameters in humans and mice. Table 2 lists the statistical results for gait parameters regression models. Gait speed decreases nonlinearly with age (ie, age²) in both species (humans: p < .0001; mice: p = .0265) and the rate of gait speed decline is not statistically different between humans and mice (species × age² interaction, p = .1665). However, the LOWESS fit curves for gait speed (Figure 1A and B) show qualitative differences. Further analysis by age group shows a marginally significant gait speed decline in younger mice (p = .0885) but not in younger humans, while gait speed declines significantly in older humans (p < .0001) but only marginally in older mice (p = .0736). Interestingly, step length and cadence (ie, two major parameters contributing to gait speed) decline differently by species (species × age² interaction; step length: p < .0001, and cadence: p = .0265). Further analysis by age group shows that step length shortens significantly in older humans (p < .0001) but not in mice, while cadence decreases significantly in older mice (p = .0226) but not in humans.

Table 2.

Regression Results for Gait Parameters for the Full Age Range, and by Age Group, for Humans and Mice

Full Age Range		Gait Speed		Step Length		Cadence
Species × age2	p	.1665		<.0001		.0265
Age2	p	.0003		<.0001		.0045
Species	p	.2268		<.0001		.6308
By age group		Younger	Older	Younger	Older	Younger	Older
Age (humans)	β	NS	−0.0018	NS	−0.0019	NS	NS
	p	NS	<.0001	NS	<.0001	NS	NS
Age (mice)	β	−0.0026	−0.0034	NS	NS	NS	−0.6016
	p	.0885	.0736	NS	NS	NS	.0226

Note: Effect of age² is tested for the full age range model due to observed nonlinear LOWESS fit curves (Figure 1). Effect of age is further tested by age group (Younger: <65; and Older: >65 maximal life expectancy) for each species.

NS = Nonsignificant; β = Standardized coefficient; p = p value (marginal significant: .05< p < .10; significant p < .05).

Discussion

While we confirm the findings of a prior study noting a shared effect of age on gait speed decline in humans and mice (1), we found that gait speed declines nonlinearly with age, with a suggestion that gait speed decline may appear earlier in mice than human males. We believe this is the first report that there are species differences between the two core gait parameters that contribute to the decline in gait speed. In human males, the decline in gait speed was primarily due to a spatial parameter (shortened step length), while in male mice the decline was due to a temporal parameter (decreased cadence). These declines are most obvious at age greater than 65% of maximal life expectancy.

Our finding on step length decline with age in humans is consistent with the previous observation (21). While gait characteristics in bipedal versus quadrupedal animals might partly explain our findings, prior observations suggest that this is not the only explanation. For example, stride length shortens with age in the rat, another quadrupedal animal (24,25). Prior evidence regarding age-related step length shortening in B6 mice is inconsistent. One study reported that stride length (eg, twice step length) does not change with aging in female mice (14), while another study reported an age-related shortening in rear-foot stride length in female but not in male mice (15). Methodological differences may explain the inconsistency between these two studies; one assessed gait on a treadmill (15) while the other did not. The treadmill fixes gait speed and treadmill gait can differ biomechanically from over-ground walking (17). Other methodological differences include how step or stride length was measured: Fischer et al. (15) used rear-foot stride length while Fahlström et al. (14) used the length between the fore- and rear-foot. Thus, age-related step length shortening appears to differ across species, and this effect is not fully explained by bipedal versus quadrupedal gait.

We found that the decrease in cadence differs across species; we found altered cadence with age in male mice but not in male humans. Although cadence is one of the two major components of gait speed, age-related cadence changes have not been previously reported for aging rodents (14–17,24,25). The ability of young mice to adjust cadence rather than step length has been observed previously only under forced changes in speed on a treadmill (23), and this adjustment in cadence due to a speed change may not be the same as cadence adjustment with age.

Our results provide a more in-depth characterization of age-related changes in gait components across these two species. Differences in gait components by species might be important for testing interventions to delay gait decline. Our study has several notable strengths. First, we measured not only speed but also its two main components, step length and cadence, to examine how spatiotemporal parameters contribute to age-associated changes in gait speed across species. The oldest B6 mice we tested are older than prior studies; our oldest mice were aged 33.3 months compared with 26 months and 28 months in prior studies (14,16). Performance-based measures in very old mice have important translational value since very old humans experience the greatest decline in physical function. We also refined our analysis to better account for species differences in body size using nondimensional normalization of the gait parameters. By using relative quick cross-sectional samples, we were able to profile age-related changes in gait parameters across species that may inform hypotheses formulation for future more complex cohort studies. Finally, the relatively large sample size and age range in our mouse study, along with the use of overground gait analysis, attempts to unveil natural walking features not previously understood before due to methodological discrepancies.

The current study has limitation; it is cross-sectional, only included males and the age distribution of the mice was skewed. Longitudinal studies could possibly yield differences due to survivor biases and should be conducted in the future. Future studies should also measure and account for relevant factors that may affect gait performance such as physical activity level or body composition. With the inclusion of males and females in large ongoing longitudinal studies at the NIA Intramural Program, we will be able to extend our work to both sexes and to within-subject change over time, as well as to a greater range of strains of mice.

This work should contribute to the future design of translational research strategies focused on functional outcomes in model organisms during aging. A better translation of interventions that lead to the preservation of function and health late in life is of critical importance. We urgently need to develop predictive biomarkers for health and survival across species in order to accelerate the translation and development of interventions that have the potential for preserving good function and health late in life. Predictive biomarkers that are developed and validated across species, such as the age-related decline in gait speed, can be used to explore the effect of emerging pro-longevity interventions. These cross-species predictive biomarkers, with their similarities and differences, can inform the extent to which preclinical findings could be extrapolated and validated in humans.

Funding

This study was supported by the Intramural Research Program of the National Institute on Aging, National Institutes of Health.

Conflict of Interest

All authors do not have conflict of interest to declare.