Terrestrial locomotion characteristics of climbing perch (Anabas testudineus) Free

Efficient walking on land is an ordinary and essential ability for extant terrestrial vertebrates. The transition from water to land is a seminal historical event in vertebrate evolution (Clack, 2002; Ashley-Ross et al., 2013). During this period, the requirements for elementary terrestrial activities (locomotion, predation and migration) posed various challenges to fluid-oriented physiological systems of earlier vertebrates. Revealing the evolutionary sequence of specialized limb structures (Wagner and Larsson, 2007) and neuromuscular mechanisms (Gillis and Blob, 2001; Lutek et al., 2022) used to coordinate these body parts is crucial for understanding vertebrate terrestrialization. The discovery of fossils (Shubin et al., 2014) and footprints (Ahlberg, 2019) belonging to the stem tetrapods has greatly enhanced our understanding of this process. These archaeological findings have allowed hypotheses regarding locomotor ability and motor neural structures to be tested by robotics methods (Nyakatura et al., 2019; Ramdya and Ijspeert, 2023). In addition, several amphibious fishes (Axlid et al., 2023; Bressman et al., 2019; Gibb et al., 2011, 2013; Perlman and Ashley-Ross, 2016; Rossi and Wright, 2021) that exhibit terrestrial locomotion with aquatic body structures have been studied, although they are not relatives of modern tetrapods. Adjusting the relative angle of the lateral surface of the tail to the ground is essential for establishing proper interactions with the substrate during the tail swing process, which poses a challenge for these amphibious fish. The tail-twist (Hsieh, 2010) or random side-lying posture (Gibb et al., 2013) enables adjustment of the relative angle between the caudal peduncle and ground, resulting in proper body–environment interactions. In particular, the specific utilization of body structures in these animals highlights the locomotor variability and the potential of non-terrestrial bodies, which could provide unique insights into the water-to-land transition.

Recently, the excavation of ancient climbing perch (Eoanabas) fossils on the Tibetan Plateau has provided evidence for a biotic dispersal bridge between Asia and Africa (Wu et al., 2019). This palaeobiogeographical route, shared with snakehead fishes, indicates the potential terrestrial locomotor ability of Eoanabas. The anatomy of extant climbing perch (Anabas testudineus), including its locomotor structures and air-breathing organ, the labyrinth structure (Tate et al., 2017), is similar to that of Eoanabas. Factors such as hunger, food competition and poor aquatic conditions significantly increase the frequency of climbing perch out of water (Pavlov et al., 2021). Moreover, their ability to move on level or sloping surfaces on land demonstrates their capacity to overcome specific land obstacles and reach new aquatic environments (Davenport and Matin, 1990). Over 200 fish species are capable of moving on land (Wright and Turko, 2016). By substituting gill cover for appendages in terrestrial locomotion, climbing perch exhibit a unique locomotor structure configuration among extant amphibious fishes. It is intriguing that climbing perch can achieve terrestrial locomotion similar to the axial–appendage-based mode without the involvement of the lateral limb, as noted but not quantified (Davenport and Matin, 1990). The rigid gill cover of the climbing perch connected to the posterior margin of the suspensorium resembles a unidirectional hinge with a restricted angle. However, little is known about how non-limb anatomical structures with restricted joint rotatability are organized for effective terrestrial movement, especially how gill cover with uniaxial rotational freedom contributes to land movement with axial undulation in climbing perch.

Coupled with the terrestrialization process, the utilization and evolution of appendages are crucial factors for diverse and effective land movement (Rackoff and Panchen, 1980; Wagner and Larsson, 2019). Previous studies have focused on the morphological modification (Shubin et al., 2006) or development (Standen et al., 2014) of appendicular skeletal-muscular structures to infer the link between morphology and corresponding motor function. Significant changes in the mechanical properties and morphology of appendages facilitate the emergence of digits and joints (Schneider and Shubin, 2013), which is considered a milestone during the fin-to-limb transition and contributes to innovative terrestrial locomotion. Joint rotatability is a key indicator of limb performance, and specific limb rotations contribute to corresponding locomotor functions, such as lifting the body and generating propulsive force (Manafzadeh et al., 2021; Molnar et al., 2021). Therefore, a feasible locomotor mode and ability are highly related to locomotor configuration, which includes structural morphology and the degrees of freedom. In contrast to flexible fins (Standen et al., 2016) or terrestrialized limbs (Chong et al., 2021), which allow complex and variable locomotion modes, the terrestrial locomotor configuration containing well-defined uniaxial rotating gill cover and axial undulation is unique but simple among amphibian fishes. Properly coordinating different locomotor body parts is essential for effective terrestrial locomotion (McInroe et al., 2016). However, the coordination between gill cover and body axial undulation for effective terrestrial locomotion remains unknown, thus hindering our understanding of the potential locomotor capabilities of this locomotor configuration.

Axial–appendage-based terrestrial locomotion requires different body structures to act in specific ways. Unlike axial or appendage-based terrestrial locomotion, the various modes of appendage-to-axial body coordination result in diverse locomotion among species (Pace and Gibb, 2014). However, relying only on brief or qualitative descriptions makes it difficult to reveal the underlying coordinating mechanism. Although it is difficult to imagine the complex motion that a living organism is capable of, animal locomotion primarily involves the coordinated actions of multiple muscles (Wainwright et al., 2008), and different muscle activation strategies enable the variability in body kinematics to perform various motion behaviours with different performances and demands (Gillis and Blob, 2001; Jayne and Lauder, 1994). Convergence in locomotion behaviour, a kinematic characteristic, suggests the potential existence of a basic motor-generating mechanism, which may be shaped by the musculoskeletal system and neuromuscular mechanisms (Catavitello et al., 2018). Searching for such simplified locomotor descriptions provides a window for investigating the basic coordinating mechanism that supports diverse and complex biological behaviours (Stephens et al., 2011). Recently, dimensionality reduction methods, such as principal component analysis (PCA) (Daffertshofer et al., 2004), have been successfully employed to quantify the complexity of natural movement at different scales (Berman, 2018). These methods help decompose large behavioural repertoires into basic coordinating units that may exist at the neural, dynamic and kinematic levels in vertebrates and invertebrates (Flash and Hochner, 2005). Such a modular hierarchy could be reused in novel ways because of its ability to learn new skills and adapt quickly to new environmental conditions (Giszter, 2015). Extracting motion characteristics from kinematic data is a necessary step towards building our understanding of how to coordinate multiple locomotor structures and gaining mechanistic insights into the musculoskeletal and neurological system.

In this study, we focused on quantitatively analysing the land movement of climbing perch. By searching for terrestrial locomotion characteristics, we explored the coordinating mechanisms and physical basis behind kinematic output and provided an insight into the motor potential of locomotor configurations containing a typical aquatic muscular system and limb-like structure with restricted mobility.

In this study, five adult climbing perch Anabas testudineus (Bloch 1792) (total length 8.5–11.2 cm, mean 9.1 cm) were obtained through the pet trade and housed in individual 50 l aquariums at 26°C. All experiments performed were approved by The Institutional Animal Care and Use Committee at Tongji Medical College, Huazhong University of Science and Technology, China (IACUC 2020-S2412). Six circularly distributed synchronous industrial cameras (The Imaging Source, DMK 33UP1300) were used to film individual climbing perch on a 1×1 m flat glass plate covered with a foam substrate at 1280×1024 pixel resolution at a rate of 100 frames s⁻¹. Before the experiment, an opaque moist cloth was placed over the fish's eyes to keep it calm for attaching marking points to its body. The individual was returned to the water for 30 min before the filmed experiments. In each trial, a fish was placed in the centre of the experimental platform and allowed to freely move for 5 min or until it left the viewing range of the cameras. The fish was returned to the water for at least 5 min between each trial. Three trials were run per fish.

Twenty-one white waterproof stickers with black circles as the markers distributed on the fish body were used to capture the body’s spatial movement and measure the locomotor structure motion, as detailed in Fig. S1B. To ensure the stability of the marking points during the experiment, it is crucial to take precautions against the interference of fish body mucus; this can be achieved by initially using a dry fibre towel to absorb surface moisture and subsequently cleaning the attachment area with alcohol. Moreover, when selecting the placement of the marking points, it is advisable to minimize crossing scales to mitigate potential moisture-related issues. The DLTdv digitizing tool (Hedrick, 2008) was used to calculate the 3D coordinates of the markers from the recorded multi-view films. The four marks on the head construct the head coordinates xyz. By measuring the positional changes of the head coordinates in the global coordinates XYZ, the movement of the entire body [p_x, p_y, p_z] was obtained (a detailed definition of the coordinates is provided in Figs S1 and S2). The roll angle θ_x describes the rotation of the body around the x-axis, the yaw θ_z around the z-axis and pitch θ_y around the y-axis. The relative angle between the actual plane of three markers fixed on the gill cover and their initial plane before the start of the movement was used to determine the joint angle of both side gill covers [θ_l, θ_r]. The markers on the body that were always visible in a trial were chosen as the points for calculating the body bending curve because some markers on the body were occasionally invisible on camera. In addition, 21 equidistant segments were obtained after interpolating the six markers on the side of the body using a cubic spline. To parameterize body undulation, the angle of each segment P_i relative to the previous segment P_i−1 was regarded as the joint angle in the current arc length position θ_i=arccos[(P_i×P_i−1)/(|P_i|| P_i−1|)] (Fig. S1).

The initiation of body swinging by an individual signified the beginning of this step, and the step was deemed complete when the individual restored an upright posture after a leap forward. The continuous terrestrial movements of climbing perch could be divided into several step cycles. The trials in which makers were visible in the six-camera surrounding view and with at least three step cycles were selected for analysis after all experiments were recorded, yielding a total sample size of 11 trials containing 34 step cycles from four individuals. To describe the spatial movement of the fish and the motion of the locomotor structures for each trial, we established four kinematics matrices, including the undulation matrix c_k(t)=[θ_l(t), … θ₂₀(t)], gill cover motion matrix g_k(t)=[θ_l(t),θ_r(t)], spatial displacement matrix p_k(t)=[p_x(t),p_y(t),p_z(t)] and spatial rotation matrix r_k(t)=[θ_x(t),θ_y(t),θ_z(t)], where k denotes the trial.

The correlation coefficients between pairs of the kinematic variables of undulation and gill cover were chosen to measure the degree of synchronization of the locomotor structures. Then, the distance of the correlation coefficients was used to identify the coordinated kinematic relationships among locomotor structures by hierarchical cluster analysis. The cluster dendrogram was constructed using the average linkage in MATLAB 2020.

When comparing the differences in motor performance (specifically displacement and rotation along the x-, y- and z-axes) across different stages, it is crucial to consider the non-independence among data from different individuals. Therefore, motion stages are defined as fixed effects and individuals as random effects (Thorson and Minto, 2015), and a linear mixed-effects model was employed to obtain estimates and confidence intervals for the fixed effects. Subsequently, one-way ANOVA was used to process the results of the linear mixed-effects model; that is, the estimated values and confidence interval (CI) of motor performance in different motion stages were used to assess the differences in motor performance across different motion stages. A probability level of P<0.05 was chosen as the level of statistical significance for each test.

We observed multistage terrestrial locomotion in climbing perch, as shown in the motion sequence diagram in Fig. 1A (see also Movie 1).

Initially, the tail and head swing to one side quickly from the upright posture. After the open gill cover is inserted into the ground, the tail quickly swings to the other side. Then, the body rolls around the anteroposterior axis so that the lateral body surface touches the ground. At this moment, a side-lying posture is adopted. Therefore, the gill cover, the side-lying body and the ground form a triangle. Then the tail swings against the ground, causing the trunk to swing around the gill cover. Lastly, the fish leaps forward due to body swing and lifting.

The body bending curves in a single step cycle (Fig. 1C) show transfer from a C-shape (frame 56) to an S-shape (frame 87). The change in contact state between the gill cover edge and the ground alters how the body interacts with the ground. Thus, a single terrestrial step cycle could be divided into three stages, and the spatial motion in each stage is shown in Fig. 2.

Generally, the total displacement of the body averaged 0.35±0.05BL (mean±s.d.) in 0.61±0.09 s in such a step cycle. Different locomotor performance characteristics emerged in the three locomotor stages. At the preparation stage, the head yaws around the vertical axis of the body 36.31 deg (95% CI, 20.67 to 51.94 deg) with the same bending direction as the tail. The head exhibits slight rolling of 9.36 deg (95% CI, −1.90 to 20.62 deg) so that the gill cover is inserted into the ground. While the gill cover supports the body, the tail swings to the other side, driving the body to rotate 34.19 deg (95% CI, 22.02 to 46.37 deg) around the anteroposterior axis further, with the entire body raised in the rolling stage. The head posture recovers to the axis of maximum displacement by the reverse yaw of 43.69 deg (95% CI, 27.95 to 59.42 deg). The lateral surface of the body is brought closer to the ground by further rolling 42.58 deg (95% CI, 31.50 to 53.66 deg). Next, the tail interacts with the ground, causing the fish to move forward in the leap-forward stage, and the gill cover is separated from the ground. In this process, the largest displacement of 0.28BL (95% CI, 0.18 to 0.39BL) is generated as a result of conversion from the gravitational potential energy stored in the lifted body at 0.05BL (95% CI, 0.03 to 0.07BL) to forward kinetic energy. The detailed performance characteristics in the different stages are listed in Table S2.

Multistage terrestrial locomotion implies that coordination of the gill covers and body contributes to various kinematic functions including adjusting the body posture relative to the ground and spatial displacement in different directions. We identified the coordination relationships among the locomotor structures using the hierarchical clustering method, where the correlation coefficient between joint angles was used to measure the degree of simultaneous motion. Using a similarity threshold of 0.6, we found four main disjunct coordinated motion groups, as shown in Fig. 3.

The anterior (body joints 1–14), posterior (body joints 15–20), and left and right gill cover were grouped separately. This result indicates that the movement of gill covers is kinematically independent of the axial movement of the body during land movement. From the single step cycle (Fig. 1), the gill cover remains fully open after being anchored on the ground. Then, the tail swings to adjust the posture of the body and lift the body. Although the gill cover does not move synchronously with the axial undulation, this suggests coordinated locomotion in which the gill cover serves as a simple anchored pole that enables axial undulation to contribute to terrestrial locomotion.

The superficial description of body bending shown in Fig. 1B is not enough to explore the characteristics of axial undulation. To quantify the characteristics of the complex axial undulation, we applied PCA to the digitized body bending joint variables (n=4, N=11 trials).

The first two principal components accounted for the variance in axial undulation explained by 87.22±3.69% (Fig. 4A). The numerical results of the first four principal components, the proportions of the simultaneous motion of the anterior-to-posterior body section, are listed in Table S1. The feature shapes that illustrate the principal components are shown in Fig. 4B, which are ordered by explained variance. The first principal component represents a C-like shape and accounts for 62.77±8.16% of the variance. The second shows an S-like shape, accounting for 24.45±7.40% of the variance. By regulating the ratio of the first two feature shapes, the motor repertoire θ*=[a¹,a²][u¹,u²]^T is generated, as shown in Fig. 4C.

As the primary area of interaction with the environment, the lateral surface of the tail converts the lateral bending force into a forward thrust force in water (Muller and Leeuwen, 2006). However, the uneven distribution of media (air–solid interface) on land poses a challenge for amphibious fish, especially in terms of converting the axial body undulation into terrestrial propulsion. In the fully aquatic fishes Danio rerio and Gambusia affinis (Gibb et al., 2011), the side-lying posture with the caudal peduncle parallel to the ground allows the lateral swing of the tail to directly interact with the ground and contributes to terrestrial tail-flip jumps. It is important to note that the tail-flip is unlikely to be used for direction-defining movements, as its overturning rotation of the head during movement prevents the fish from visually perceiving the end of its movement. Additionally, maintaining a defined direction before and after the jump is difficult in the side-lying posture. The trend of a slender body in amphibious fish (such as anguilliform or eel-like body form) facilitates land invasion by increasing the contact point with the ground for forward propulsion, especially in complex 3D habitats (Ward et al., 2015). Another alternative observed in terrestrial blennies (Alticus arnoldorum) (Hsieh, 2010) is twisting the tail to mediate the orientation of the caudal peduncle for maximally turning the lateral bending force into directing propulsive forces against the ground. This tail-twist ability enables greater locomotion than lateral side-to-side motion in amphibious blenny (Praealticus labrovittatus) (Hsieh, 2010). In our study, we characterized novel terrestrial locomotor abilities in climbing perch, which involve active rotation of the entire body around the anteroposterior axis. The gill cover plays a crucial role in anchoring the body to the ground, thus allowing body undulation to actively adjust its posture from upright to side-lying and eventually recovering to upright at the end of the leap-forward stage. Significant displacement occurs during the leap-forward stage, benefiting from adjustments in body posture for proper interaction with the ground. In contrast to fully aquatic fishes with uncontrolled side-lying postures caused by random body bounce, this unique terrestrial reorientation ability exhibited in climbing perch suggests the potential for directed terrestrial locomotion.

Compared with the large rotation ability in flexion/extension, adduction/abduction and pronation/supination that are common in modern tetrapods, the gill covers of climbing perch exhibit adduction/abduction only. Connected to the posterior margin of the suspensorium with a membrane, as shown in Fig. S2C, the rigid gill cover functions as a simple pole with an angle-limited hinge (Davenport and Matin, 1990). When the gill cover is inserted into the ground, the reverse support force from the ground to the end of the gill cover allows the gill cover to maintain the maximum opening angle under the passive tension of the flexible membrane. Similarly, the pectoral girdle of Polypterus is joined to the skull via the post-temporal bone so that the pectoral fin can support the body with a stable connection (Standen et al., 2014), and the fin muscles prevent fin collapse, add stabilization and increase body support (Foster et al., 2018). Compared with the burst power demands on the muscle near pectoral girdles for the stability of fins (Du and Standen, 2017), gill cover with a maximum opening angle restricted by biological structural constraints could reduce the demand for muscle force and provide stable support for the terrestrial locomotion of climbing perch. Further analysis of the mechanical mechanism of gill cover will help us explore the motor potential of the locomotor configuration in climbing perch, as well as that of animals with similar locomotor configurations.

The proper synchronous motion of the appendages and body enhances terrestrial locomotor performance in mudskipper (McInroe et al., 2016). The coordination of gill cover with axial undulation, such that the inertial and interaction forces of axial undulation enable terrestrial reorientation and propulsion, has not been well studied. Our quantitative research was conducted to examine the coordination characteristics between gill cover and body undulation. The results of the cluster analysis showed kinematic independence between the motion of the gill covers and the body. According to the kinematics of a single step cycle (Fig. 1A), the gill cover maintains its maximum opening once it is anchored to the ground, and the axial undulation drives the body to flip forwards. Synchronized movements of different body parts imply that the muscles controlling these body parts are activated synchronously by specific neural circuit structures. The coordination of gill cover with axial undulation described here reduces the requirement for synchronicity between those body parts.

Typically, body undulation contributes to swimming, while gill cover assists fish in obtaining oxygen in water. The collaborative utilization of these two body structures, which function differently in water, to achieve efficient land locomotion is truly remarkable. Aquatic or amphibious vertebrates often require the reuse or modification of swimming neural–muscular mechanisms to regulate the coordination of different movement structures for terrestrial locomotion (Lutek et al., 2022). The swimming central pattern generator (CPG) with extension of the limb oscillator centre reveals a potential coordinated motor-generating mechanism for salamander walking on land (Ijspeert et al., 2007). The direct interconnections between the neurons that regulate the muscles engaged in the movement of diverse biological structures, as observed in CPGs, dictate synchronized muscle activation, resulting in tightly coordinated movements (Grillner and El Manira, 2019). The opening and closing of the gill cover are manipulated by the opercular muscles, including the levator, dilatator and adductor operculi muscles, which also assist continuous air ventilation with active exhalation in Anabas (Liem, 1987). The adductor and levator operculi muscles are considered hyoid musculatures and are innervated by the VIIth nerve connected to the medulla of the hindbrain, which contains numerous sensory neurons. The dilatator operculi muscle, as a mandibular musculature, is innervated by the Vth nerve emerging from the lateral aspect of the metencephalon (Butler, 2000). The CPGs in the spinal cord receiving descending projections from brainstem reticulospinal neurons mediate limb and body oscillatory centres to generate coordinated movement (Ryczko et al., 2020). In contrast to the CPG neural circuit with direct interconnected neurons, the neurons controlling gill cover and body axial undulation are connected in different areas of the brain. Therefore, the difference in the neural descending pathways that regulate the motion of the gill cover and body undulation separately may explain the kinematic independence of the movement of gill covers and body undulation.

Axial undulation serves multiple functions, including actively altering the terrestrial orientation and producing propulsive force. Our research on the motion characteristics of axial undulations shows that complex undulations are reconstructed as a combination of a few feature shapes. The first two feature shapes explained more than 87% of the variance in the axial undulation. The visualization of those feature shapes shows that the first feature shape represents an undulation characteristic of the body bending uniformly to one side, while the second feature shape shows reverse bending in the anterior and posterior body. However, the physiological basis of such characteristic movements has never been mentioned before. The axial undulation is powered by stacked myomeres separated by myosepta, the arrangement of which remained unchanged throughout 400 Myr of gnathostome evolution (Gemballa et al., 2003). The lateral muscle system of climbing perch was consistent with that of typical bony fish (Morioka et al., 2008). Considering the natural characteristics of the axial muscle structure of fish, the lateral body deformation patterns that correspond to motion characteristics may be relevant to muscle activation strategies modulated by specific spinal neural circuits (Altringham and Ellerby, 1999). The neural circuits underlying muscle activation patterns, including the swimming-CPG and fast-escape neural circuits, have been identified in teleost fish (Berg et al., 2018). The two primary axial undulation neural circuits in water have different neural connection structures and activate muscles in different ways, yet they yield kinematic outputs that resemble a C-shaped or S-shaped movement (Liu and Hale, 2017), similar to the extracted feature shapes of terrestrial locomotion in climbing perch. Elicited by environmental transition, alterations in muscle activation may mediate the functional role of major locomotor muscles and underlie the ability of many species to move through different environments (Gillis and Blob, 2001). Although it is difficult to discern the neural circuit basis of terrestrial locomotion in climbing perch via kinematic analysis, this provides a starting point for identifying neuromuscular mechanisms. Further, EMG signal analysis will help us identify neural circuits that are modified or reused in the terrestrial locomotion of climbing perch, which will contribute to our understanding of motor performance and limitations.

The terrestrial locomotion of climbing perch also provides an alternative model for investigating the movement potential of animals with similar locomotor configurations containing a typical aquatic lateral muscular system and lateral structure with restricted rotatability. Importantly, the identification of alternative locomotor principles as demonstrated here for climbing perch will inspire biomimetic design, which should be helpful for further examining the function of locomotor structures and the motor potential of motion-generating mechanisms, and for gaining new insight into evolution and neurobiology.

We thank Chuang Liu for his assistance in the animal motion capture experiments, Dr Quanlin Li for his help in setting up the kinematics capture platform, and Professor Feixiang Wu for providing valuable information about the physiological characteristics and habitat of the climbing perch.