Wall following in Xenopus laevis is passive

The tendency of animals to follow boundaries within their environment can serve as a strategy for spatial learning or defence. We examined whether animals of Xenopus laevis employ such a strategy by characterizing their swimming behaviour. We also investigated potential developmental changes, the influence of tentacles, which some of the developmental stages possess, and whether wall-following is active (animals seek out wall contact) or passive. Animals’ swimming movements were recorded with a camera from above in a square tank with shallow water and their trajectories were analysed especially for proximity to the nearest wall. With the exception of young larvae, in which wall following was less strong, the vast majority of animals – tadpoles and froglets – spent more time near the wall than what would be expected from the proportion of the area near the wall. The total distance covered was not a confounding factor. Wall following was also not influenced by whether the surrounding of the tank was black or white, illuminated by infrared light, or by the presence or absence of tentacles. Animals were stronger wall followers in smaller tanks. When given a choice in a convex tank to swim straight and leave the wall or turn to follow the wall, the animals consistently left the wall, indicating that wall following in Xenopus laevis is passive. This implies that wall following behaviour in Xenopus derives from constraints imposed by the environment (or the experimenter) and is unlikely a strategy for spatial learning or safety-seeking. Summary statement: Xenopus laevis tadpoles and froglets tend to swim along the walls of a square tank; but this wall following is passive – in a convex tank, they leave the wall.

experiments. The main difference was that in some cases, background subtraction was carried out before 1 3 9 thresholding the image, whereas in other cases images were thresholded directly, either using a simple or a Gaussian In contrast, the following steps applied to all cases. The contours of the animals were extracted and the largest 1 4 2 contour was taken as the animal. X-Y positions were then calculated relative to the tank geometry. This position were generated to ensure that the animal was tracked faithfully. Erroneously tracked frames were identified 1 4 6 by visual inspection and spuriously high forward velocities, and their X-Y coordinates were interpolated. Such 1 4 7 6 corrections were necessary in 36 video sequences, 22 of which were animals in the standard condition, with 1 4 8 maximally 16 frames to interpolate. In some cases, none of the tracking strategies proved successful, leading to an 1 4 9 exclusion of 9 animals in the standard condition. From the X-Y position in the tank-warped images, parameters such as the distance covered during the 1 5 2 swimming and the distance to the nearest wall were calculated. To avoid including jitter as animal movement, the 1 5 3 trajectories were simplified with the Ramer-Douglas-Peucker algorithm (using the rdp python package, 1 5 4 https://github.com/fhirschmann/rdp). The epsilon parameter, which determines the degree of simplification, was set 1 5 5 to 10 in a 900 x 900 pixel video, and was scaled linearly to adjust for changes in the resolution. The simplified 1 5 6 trajectory was then used to calculate the total distance the animal covered during swimming. Only animals that 1 5 7 covered a distance of at least one side length of the tank were included in the analysis; in the standard condition, this proportion of time that the animal spent near the wall was calculated. While it is desirable to keep the 'near wall' 1 6 0 threshold as small as possible, 15 mm was chosen to ensure that the tracked centroid of the large animals was still 1 6 1 within that threshold when the animal was near a wall. With 15 mm, the 'near wall' area constituted 29.1% of the 19 1 6 2 x 19 cm tank. When comparing different tank sizes (7 and 19 cm side length), the animals were compared with a 15 mm 1 6 4 'near wall' threshold -which might indicate the attractiveness of the wall independent of the size of the tank. However, since the 'near wall' area in the 7 x 7 cm tank constitutes 67.3% of the whole tank, the distribution of 1 6 6 distances to the wall in both tanks were normalised to the maximum distance, and a threshold was chosen to define 1 6 7 the 'near wall' area as intermediate in the proportion between the 29.1% and 67.3% that resulted from the 15 mm 1 6 8 threshold. Therefore, 0.28 of the maximal distance from the wall was chosen as a threshold for defining the 'near wall' area independent of the tank's size, yielding a 'near wall' area of 48% in both tanks, which was intermediate  The python code used to analyse the data and the tracked data can be found on figshare (Hänzi and Straka,  Parameters of interest were tested for normality using a Shapiro-Wilk test; the appropriate parametric or non-1 7 6 parametric tests were chosen accordingly, using an alpha value of 0.05. The distribution of the proportion of time 1 7 7 spent near the wall of all animals in the standard condition was not normally distributed; therefore Spearman rank Systems Incorporated, San Jose, USA). The swimming behaviour of animals in a square tank between pre-metamorphic stage 47 (larvae) and post-1 8 4 metamorphic stage 66 (froglets) was quantified by monitoring the animals' trajectories over a period of 10 minutes 1 8 5 in each individual (Fig. 1). Examples of animals at different developmental stages revealed a variety of swimming 1 8 6 behaviours with respect to the walls of the tank. Independent of developmental stage, some animals exhibited 1 8 7 trajectories that appeared to cover the entire tank ( Fig. 2A-C), while others swam preferentially along the walls of 1 8 8 the tank (e.g. Fig. 2D,G). To visualise the extent of wall following, the cumulative frequency of distances to the 1 8 9 nearest wall over the 10 minutes period of swimming was plotted (see Fig. 1B). This graphical presentation is 1 9 0 equivalent to a histogram of distances to the nearest wall that are summed up along the X-axis. The cumulative frequencies of distances to the nearest wall for all animals (n = 79) are shown in Figure 3A. The proportion of time that the animals spent near the wall (within 15 mm of the wall) was taken as a measure of the 1 9 3 strength of wall following. As a group, the 79 animals differed significantly from the proportion that could be 1 9 4 expected from the 'near wall' area (29%, Fig. 3B, Wilcoxon signed rank test, p < 0.0001). Five animals, however, 1 9 5 spent less than 29% of their time near the wall, which is the proportion of the 'near wall' area. Four of these were of 1 9 6 developmental stage 48 or below and this tied in well with the impression that the strength of wall following wall, rho = 0.48, p < 0.0001, n = 79), suggesting that Xenopus larvae/froglets become stronger wall followers during  To reveal potential changes in wall following behaviour in individual animals over the 10 minute test period, the entire test period. Moreover, the total distance covered within the 10 minutes was no confounding factor for wall 2 0 5 following, since the rank correlation between the total length of the trajectory and the proportion of time spent near 2 0 6 the wall was not significant (Fig. 3E, Spearman's rank correlation, rho = 0.03, p = 0.77). of rod-like appendages that protrude from the corners of their mouths (Nieuwkoop and Faber, 1956). These 2 1 0 appendages might be necessary or at least advantageous for wall following, given the presence of Merkel cells, 2 1 1 potentially assigning a tactile function to these tentacles (Ovalle, 1979;Ovalle et al., 1998). However, contrasting tentacles. This allowed to directly test the influence of tentacles on the degree of wall following. Accordingly, the 2 1 4 swimming behaviour of a population of tadpoles at developmental stages 54 -60 without appendages (n = 11) was 2 1 5 compared with that of an age-matched group of tadpoles (n = 13) that possessed tentacles with a length of at least 3 2 1 6 8 mm.

1 7
Statistical analysis of the swimming behaviour as reported above indicated that both populations of animals 2 1 8 had a similar propensity for wall following (blue and red traces in Fig. 4A). This is demonstrated by the overlapping 2 1 9 distributions of the cumulative frequencies of distances to the nearest wall in animals with and without tentacles 2 2 0 (blue and red traces in Fig. 4A). The proportions of time that these animals spent near the wall were not significantly 2 2 1 different between animals with and without tentacles (Fig. 4B, Mann-Whitney-U test, p = 0.09). If anything, animals 2 2 2 without tentacles were located closer to the wall than animals with tentacles (see blue and red traces in inset in Fig.   2 2 3 4A). This likely derives from the fact that the presence of tentacles creates an additional distance of the tadpole with 2 2 4 respect to the wall that is not present in animals without tentacles. Tentacles are therefore no prerequisite for wall 2 2 5 following. This, however, does not exclude that tentacles are used as tactile probes; rather it shows that despite the 2 2 6 absence of tentacles, tadpoles follow the walls of a tank and potentially use facial skin areas as tactile probes. The wall following of Xenopus larvae/froglets analysed above was further examined during swimming under tadpoles/froglets (n = 10) was compared in a tank in which the four walls were covered on the outside by a white or 2 3 2 a black background (Fig. 5A,B). Analysis of the swimming behaviour indicated that the propensity for wall the wall in the two conditions (paired t-test, p = 0.59). This suggests that the visual system exerts no apparent 2 3 5 influence on the tendency of Xenopus for wall following. This conclusion was confirmed by another set of rank test, p = 0.47), indicating that the reduced light condition during infrared illumination had no effect on wall 2 4 0 following.

4 1
Influence of tank size on wall following 2 4 2 Wall following might be influenced by the size of the environment. To test whether the wall is equally that the animals spend more time near the wall in the smaller tank (Fig. 6A). This is confirmed by comparing the in the small tank are significantly larger (Fig. 6B, paired Wilcoxon signed rank test, p = 0.0078). This suggests that 2 4 8 the wall is more attractive in the smaller tank. However, the 'near wall' area (within 15 mm of the wall) is also 2 4 9 relatively larger in the smaller tank (67.3% of the total area in the 7 x 7 cm tank vs. 29.1% of the total area in the 19 2 5 0 x 19 cm tank). To compare wall following on the same scale, the distances to the wall were normalised to their 2 5 1 maximum, and a threshold was chosen that resulted in an intermediate 'near wall' area (threshold of 28% of the 2 5 2 9 maximal distance to the wall, resulting in a 'near wall' area of 48% of the total tank area; Fig. 6C). The proportion 2 5 3 of time spent in these area-normalised 'near wall' areas was again significantly larger in the smaller tank (Fig. 6D, paired Wilcoxon signed rank test, p = 0.0078). The animals are therefore stronger wall followers in the smaller tank 2 5 5 also when taking into account the differences in area.

5 6
Wall following is passive 2 5 7 Wall following might be either active such as in blind cavefish (Patton et al., 2010) or passive (distinction 2 5 8 according to Creed and Miller, 1990). To distinguish between the two possibilities for wall following in larval and 2 5 9 adult Xenopus, the swimming behaviour was tested in a specifically designed tank (Fig. 7A,B). The use of a tank in 2 6 0 which two of the four walls had convex curvatures allowed testing if tadpoles seek wall touch during swimming 2 6 1 actively or follow concave walls passively (red and blue arrows in Fig. 7A). The proportion of trials when animals 2 6 2 swam straight after encountering a convex curve (Fig. 7B) was evaluated from visual inspection by the experimenter. The majority of tested tadpoles swam straight in all trials (Fig. 7B) more or less independent of their 2 6 4 developmental stage (Fig. 7C,D, n = 22), leading to the conclusion that wall following in Xenopus is passive. Xenopus laevis -from small tadpoles to froglets -tend to follow the wall when swimming in a square tank. The strength of wall following increases with progressive development and smaller tank size and is not confounded does not lead to stronger wall following compared to animals that naturally do not develop these appendages. Also, 2 7 1 vision is unlikely a main driver of wall following, as surrounding the tank by black or white paper or changing the 2 7 2 light to infrared illumination does not change the strength of wall following. Wall following is passive as indicated 2 7 3 by straight swimming in a tank with convex curvatures. This indicates that wall following in Xenopus is likely imposed by the concave environment. Wall following being passive might also explain why it persists across 2 7 5 metamorphosis and is present in both tadpoles and froglets, independent of their very different locomotor styles. Classification and different types of wall following 2 7 7 Wall following in concave environments has been described for a wide variety of animals: from Martin, 2004), or cockroaches (Camhi and Johnson, 1999;Jeanson et al., 2003;Okada and Toh, 2000), to fishes 2 8 0 such as zebrafish (Anichtchik et al., 2004;Colwill andCreton, 2011), goldfish (Kato et al., 1996), salmon (Clements  (Patton et al., 2010;Teyke, 1985;Teyke, 1989), to several rodent species including 2 8 2 voles, rats and mice (Eilam, 2004;Perrot-Sinal et al., 1999;Simon et al., 1994;Treit and Fundytus, 1988;Webster et 2 8 3 al., 1979;Wilson et al., 1976). In many cases, these examples of wall following behaviours have been described in dark than in the light (Eilam, 2004), though some authors use the term centrophobism without necessarily implying 2 9 1 a visual mechanism. Thus, centrophobism and thigmotaxis are two potential but not mutually exclusive mechanisms 2 9 2 that can lead to the avoidance of open spaces and the following of environmental boundaries. Wall following is 2 9 3 therefore a neutral term to describe the tendency of an animal to follow vertical walls in its environment without a 2 9 4 reference to the underlying mechanism. An environment with convex borders allows distinguishing between passive 2 9 5 and active wall following (Creed and Miller, 1990). Animals perform active wall following when voluntarily 2 9 6 seeking out the proximity to a wall and turn in order to remain near the wall. Passive wall following occurs when 2 9 7 animals leave the wall at a convex curve but follow the walls in a concave environment. When wall following is 2 9 8 active, thigmotaxis, centrophobism or a combination of the two can be the underlying mechanism. Potential uses of wall following 3 0 0 Thigmotaxis has been described both as a defensive strategy (Grossen and Kelley, 1972) as well as a spatial thigmotaxis and centrophobism. Independent of the underlying mechanisms, the use as a defensive strategy is clear.
Moreover, wall following can also serve as a useful spatial exploration strategy. Especially under conditions when 3 1 1 long-range sensing such as vision is not available, exploration of the environment based on touch along its borders 3 1 2 can provide the basis for the formation of a cognitive map (Kallai et al., 2007;Yaski et al., 2009) and serve as a 3 1 3 reference frame for later exploration (Kallai et al., 2005). However, this is only useful as an initial strategy; if it is 3 1 4 used excessively it can even prevent further spatial learning (Kallai et al., 2007). Such initial wall following as a 3 1 5 means for spatial learning has been observed in various species such as crayfish (Basil and Sandeman, 1999), blind  In this study we examined a range of developmental stages of Xenopus -from small to large tadpoles immediately prior to metamorphosis as well as froglets after metamorphosis has been completed. Wall following in 3 2 0 a square tank was present at all developmental stages; the strength of wall following was weakest, however, in the 3 2 1 smallest tadpoles, stronger, with considerable variations in larger tadpoles and consistently strong in froglets. This 3 2 2 persistence suggests that wall following is not a behavioural strategy only employed by tadpoles or frogs, and is not 3 2 3 11 linked to a particular locomotor style such as undulatory tail-based propulsion or leg-based swimming. Moreover, 3 2 4 wall following in a convex tank was passive in all animals tested (see below). The weaker wall following in young 3 2 5 larvae is noticeable and might be related to the somewhat different swimming style of these animals (see Fig. 3A in facilitate turns away from a vertical wall in young larvae and explain the weaker wall following. which are retracted during undulatory swimming (Hänzi et al., 2015). These tentacles -like other skin areas -3 3 3 possess mechanoreceptive Merkel cells (Nurse et al., 1983;Ovalle, 1979;Ovalle et al., 1998), and therefore the 3 3 4 tentacles likely serve a tactile function when the animal is stationary or cruising slowly with tentacles extended 3 3 5 forward. We hypothesised that these tentacles might be used to explore the environment in a way that is similar to 3 3 6 rodents' whiskers but simpler because the structure is not as specialised. However, younger larvae and older animals 3 3 7 at metamorphic climax (>stage 61) that do not possess any tentacles were overall similar in their wall following 3 3 8 tendencies, as were animals that for unknown reasons did not develop tentacles (Fig. 4). While this does not exclude that -when present -tentacles are used for tactile exploration, it shows at least that tentacles are not necessary for 3 4 0 wall following, and if tactile exploration is needed, tadpoles might also use their facial skin. open in the dark (Diaz, 1992;Price et al., 1984;Vasquez, 1996), and some authors also assign a role of vision in the avoidance of open spaces by rats (Cardenas et al., 2001;Martínez et al., 2002). However, tadpoles and froglets of 3 4 8 Xenopus laevis did not show stronger wall following in light than under infrared illumination. While the IR lamps 3 4 9 used here did not produce pure infrared light, IR illumination nevertheless is a condition with considerably reduced  the geometry of the environment has shown to influence path shapes of rats not only at the perimeter but also at the 3 5 7 centre of an environment (Yaski et al., 2011). A wall can exert both a guiding and attracting influence on mouse 3 5 8 behaviour from quite some distance (Horev et al., 2007). Two studies explicitly examined the proportion of time that    near the wall and the developmental stage of the tested animals (n = 79); note the significant Spearman's rank indicates that total covered distance is not a confounding factor for the degree of wall following as measured by the frequency distributions of the distance to the nearest wall of animals with tentacles (red, n = 13) and of animals 4 4 0 without tentacles (blue, n = 11) between developmental stages 54 -60; the inset is a higher magnification of the 4 4 1 initial part of the cumulative frequency distribution and shows that tadpoles without tentacles (blue) align closer nearest wall during swimming of stage 47 -50 tadpoles (n = 9) over a 10 min period in a 7 x 7 cm tank (green) and Wilcoxon signed rank test, p = 0.0078, n = 9). only animals with at least 4 trials were included.