MouseWalker

We previously described an approach to track and quantify kinematic properties of untethered walking fruit flies [21]. Using this approach we quantitatively described the walking behavior of wild-type and genetically modified animals [21, 22]. This method is based on the reflection of light within a transparent material through an optical effect termed total internal reflection [23, 24]. Foot contacts disrupt this effect causing fTIR, which generates scattered light that can be detected by a high-speed video camera (Fig. 1a). We built a simple walking apparatus from readily available and inexpensive supplies (Additional file 1: Figure S1), mostly precut acrylic glass and aluminum components (see Additional file 1: Figure S1, Additional file 2: Figure S2, Additional file 3: Figure S3, Additional file 4: Figure S4, Additional file 5: Figure S5 and “Methods” for details), in which the rodents can walk freely down a narrow corridor. The floor is of acrylic glass surrounded by LED lights, thus producing a touch sensor (Additional file 2: Figure S2). Empirically, we found that an acrylic glass surface resulted in a better fTIR signal-to-noise ratio compared to glass, possibly due to a rougher surface thus allowing more contact between the animal’s paws and the walking surface. Although acrylic glass scratches more easily than glass, thus interfering with the fITR signal and subsequent tracking, it can be easily replaced. A light box positioned above the walking apparatus allows the outline of the mouse body to be visualized as the animal moves along the walkway (Additional file 5: Figure S5). Finally, a mirror placed at 45° below the walking surface reflects the fTIR signal and body outline, allowing them to be captured by a high-speed camera (Fig. 1a and Additional file 4: Figure S4 and Additional file 6: Video S1). Depending on the type of camera available and color of the LEDs and light box, the setup can generate monochromatic or color videos (Fig. 1b).

Recorded videos are loaded into a program developed specifically for this assay (Fig. 2a'). The MouseWalker software was written in MATLAB and compiled as a standalone program, which analyzes the sequence of images from the videos by registering the position of the body, tail, and each footprint (Additional file 7: Video S2). Each video is loaded into the graphics user interface (GUI) of the program where the auto-tracking feature can identify each footprint, body contour, and position of the tail, with an accuracy >90 %, depending on the acquisition conditions and settings. Footprints and body contours are identified based on pixel intensity thresholds defined by the user in a dedicated settings window, which can be determined in a few minutes by an experienced user with the help of a preview section or by auto-tracking for a few frames (Fig. 2a'). Optimal settings can be stored for subsequent movies. Importantly, since the program can discriminate between red, green, and blue inputs, body elements can be identified based on color (Fig. 1b'). Subsequently, the user can manually edit any mislabeled footprints or body features if necessary. In addition, the user can toggle between different visualization modes (such as the unprocessed image, tracked body, or footprints; Fig. 2b and Additional file 7: Video S2, Additional file 8: Video S3, Additional file 9: Video S4), which help in setting the parameters and accurately editing the video. These are particularly important in identifying the footprint contacts and body. Most importantly, the user interface allows the generation of a set of output files (see “Methods” for a complete list), including an Excel file containing >20 quantifiable parameters (Additional file 10: Table S1) and an annotated video (Additional file 7: Video S2, Additional file 8: Video S3, Additional file 9: Video S4). The system is flexible in that add-on scripts can be written to extract additional parameters from the MATLAB tracks or Excel files.

Walking by wild-type mice

As a proof of principle, we examined the walking behavior of a laboratory wild-type C57BL/6J strain moving across an 80-cm-long walkway. We collected 16 videos from four animals and analyzed the data as they moved continuously in the center 50 cm of the walkway.

The footprint test is a common qualitative readout that displays the footprint pattern and footprint contacts, potentially highlighting gait abnormalities [13]. MouseWalker can generate such a readout displaying the footprint pattern generated by the walking animal in addition to the path created by the body center (Fig. 3a). Footprints are converted to heat maps, as determined by the intensities of the fTIR signal, which are proportional to the pressure applied. Using a proper calibration procedure where, for example, all four paws are in contact with the acrylic surface while the animal is immobile, the user can quantify the pressure. Our data set displays the typical wild-type pattern with evenly spaced steps. Hind paw placement partially overlaps the previous forepaw placement, although offset at slightly lateral positions. Since this footprint overlap makes leg identification more challenging, an additional footprint pattern is generated where pixel intensity is eliminated by a leg-specific color code, allowing the unambiguous identification of the footprints from each leg (Fig. 3a' and Additional file 11: Video S5).

Because video acquisition is carried out at a high temporal resolution of 250 Hz, a series of frames spanning an entire stance phase can be generated. The data extracted from this series are referred to as footprint dynamics. At 250 Hz, we observed that the initial contact with the surface at touchdown is typically done with the entire paw touching the surface almost simultaneously, with most of the pressure exerted by the metatarsal and metacarpal pads on the hind and fore legs, respectively (Fig. 3b, left section). This characteristic is distinct from human walking, where stance phases begin with contact by the heel [25], or the cat, where contact is initiated by the most rostral section of the paw [26, 27]. During the stance phase in the mouse, pressure is gradually transferred rostrally to the toes prior to liftoff (Fig. 3b right section). The time spent when the stance phase was supported by the toes can be up to one-third of the entire stance phase (Fig. 3b). Visual inspection of the complete data set suggests that this behavior is independent of the walking speed (data not shown).

The MouseWalker program also compiles the complete set of footprints present in each video (Fig. 3c). We observed that at the end of the swing phase, just prior to touchdown, the toes changed their conformation from a closed to an actively open conformation (data not shown). This behavior occurs typically 20–30 ms before contact with the ground, indicating that it is under active neuronal control. Consistent with this notion, toe spreading has been used as a metric to study sciatic nerve function (for example [28, 29]). Due to the orientation of the footprints relative to the displacement axis, automatic quantification of toe spreading can be prone to errors. Nevertheless, each set of images generated by MouseWalker is provided with a scale calibration, allowing the user to measure distances easily and accurately using ImageJ from the National Institutes of Health (NIH) or similar software.

We also quantified several gait parameters as a function of walking speed (Fig. 3d–g). We observed that faster animals display ~20 % longer step strides, changing from ~25 mm in slower animals to ~30 mm in faster animals (Fig. 3d). A much larger variation was observed in stance phase duration (Fig. 3e), with values becoming exponentially shorter as speed increases, with a minimum of approximately 50 ms for the fastest animals. Swing duration exhibits a much smaller variation as observed in other experimental conditions [30, 31]. Although a similar variation is observed in invertebrate systems, stance phases last longer than swing phases at all speeds [21, 32]. In contrast, in our data set, we observed that above ~50 cm/s, swing duration is longer than stance phase duration, consistent with recent results [33]. We also used these data to measure the duty factor, defined as the fraction of the step cycle where the leg is in the stance phase (stance duration / period) [34]. This parameter has been used to distinguish walks from runs, as values ≥0.5 are described as walks, while values below 0.5 are considered runs [35, 36]. From our data, we find that swing duration surpasses stance duration at a duty factor of 0.5, which corresponds to a speed of 52.8 cm/s (Fig. 3f). Above this speed, the feet spend on average more time in the swing phase than in the stance phase, which is typical of running [34]. Accordingly, swing speeds also increase with increased speed (Fig. 3g).

Our MouseWalker software tracks not only footprints but also body features. Despite some tracking inaccuracies of the body center and footprint center, these measurements allow the stance phase of each leg to be reconstructed as it is anchored at the floor relative to the body (Fig. 4a and Additional file 12: Video S6). Thus, each stance trace reflects the amount of body wobble during stance phases. Each stance trace is normalized to body length to account for variations in body size and is defined as the position of the foot relative to the center of the body from paw touchdown (anterior extreme position, AEP) to the end of the stance phase (posterior extreme position, PEP) [21]. Regardless of the speed, stance traces run parallel to the body axis with the forelegs positioned more medially (Fig. 4a). A measure of the straightness of the stance traces, the stance linearity index, is calculated by computing the average difference between the actual stance trace and a smoothed version of the trace [21]. Similarly to what was observed for the fruit fly Drosophila melanogaster, stance traces become straighter as speed increases (Fig. 4b). In addition, we also measured the variability in the AEP and PEP coordinates for all steps in each video (Fig. 4c). This parameter, termed footprint clustering, corresponds to the standard deviation of the average AEP and PEP coordinates for each video [21]. Thus a smaller value for this parameter corresponds to a more consistent position for paw touchdown or takeoff (AEP or PEP, respectively). As with Drosophila, we observed smaller footprint clustering values for AEP compared to PEP, indicating a more consistent foot placement at the onset of the stance phase, possibly due to tighter motor control. In contrast to the fly, where faster animals had smaller footprint clustering values, there was little dependence on speed for this parameter in the mouse [21].

Adult mice are typically described as using a walk gait at slower speeds, a trot gait at intermediate speeds, and a gallop or bound gait at higher speeds [33, 37], although additional variants have been described [36, 38]. The walk gait is generally defined as when only one leg is swinging, a trot is defined by the simultaneous swinging of diagonal feet, and a gallop has two defining characteristics: only a single foot is in the stance phase and there is an aerial phase (no feet are in the stance phase) [39]. With these definitions in mind, MouseWalker outputs the step pattern with the instantaneous speed and step combinations associated with each frame in the video (Fig. 4d). Although videos are selected based on the animal’s average speed, plots of instantaneous speed display a wave-like appearance with minimum speeds typically occurring at step transitions (Fig. 4d). These observations are consistent with other experimental conditions in mice as with other quadrupeds and walking insects [21, 31, 40, 41]. As a proxy for assessing the presence of specific gaits, MouseWalker computes the fraction of frames assigned to a particular leg combination (Fig. 4e–g and Additional file 13: Figure S6). We defined seven categories for the possible stance combinations: no swing, single-leg swing (regardless of the position), diagonal-leg swing, lateral-leg swing (both left or both right legs), front or hind swing (both hind or both fore legs), three-leg swing, or all-legs swing. In our data set, the two diagonal swing conformations, which are typically observed in the trot gait, are the most representative configuration, present in more than 50 % of the frames (Fig. 4e). The presence of this gait pattern increases with speed, reaching approximately 80 % even at intermediate speeds. Concurrently, there is a decrease in the fraction of frames in which only a single leg is swinging (Fig. 4f), typical of walk gaits, and in the fraction of frames with all feet in the stance phase (Additional file 13: Figure S6A). We also observed an increase in the fraction of frames with three legs in a swing position, typical of gallop gaits (Additional file 13: Fig. S6B). These results reveal that there is progressively reduced contact with the ground as speed increases. Consistently, for speeds greater than 52.8 cm/s, the mice have a duty factor below 0.5 (Fig. 3f), suggesting the presence of an aerial phase [34]. We find that for most of the animals with a duty factor below 0.5, and thus are considered to show a run-like behavior, there are some frames in which all legs are in the swing phase (Additional file 14: Table S2 and Fig. 4g). The MouseWalker program also calculates the fraction of lateral swing conformations, typical of the pace gait (Additional file 13: Figure S6C), and simultaneous front or hind swing, found in bound and hopping gaits (Additional file 13: Figure S6D). In our data set, we found a poor correlation between these step configurations and speed (Additional file 13: Figure S6C, D). Although most laboratory strains rarely display bound or hopping gaits [33], some mouse species and genetically modified mice that lack left–right coordination use these gaits more frequently [38, 42].

Finally, the MouseWalker software package also analyzes phase values between contralateral legs (Additional file 13: Figure S6E), which provide a measure of inter-leg coordination. As expected, these measurements indicate that contralateral legs move in anti-phase with phase values of 0.507 ± 0.041 and 0.491 ± 0.061 for fore and hind legs, respectively.