Diversification of pectoral control through motor pool extension Authors: Ruth Gutjahr, Maximilian S. Bothe, Thorin Jonsson, Boris P. Chagnaud Published: 2024 ABSTRACT SignificanceVertebrates are able to perform highly complex motor behaviors with their pectoral appendages, such as writing or throwing a ball. We show in a teleost fish that performs different pectoral behaviors, that pectoral control is enabled by a spatial segregation that allows for a coordinated recruitment of the motoneuron pools innervating antagonistic muscles. This segregation is accompanied by a size-dependent spatial organization of motoneurons and a size-dependent differential tuning of intrinsic membrane properties of motoneurons. We show that the interplay of these anatomical and physiological organization principles underlies the flexibility of pectoral control. We suggest that these principles are not a unique feature of tetrapods, but are shared across vertebrates. The versatile control of human arms and hands is an outstanding example of the flexibility of pectoral control. Since their first evolutionary appearance (1), pectoral appendages have undergone substantial anatomical and physiological adaptations, which are accompanied by varying associated motor outputs, ranging from slow to fast and from small to large amplitude motions across and within species. While the gross anatomical architecture and the motion patterns of pectoral appendages have been investigated in numerous animals (for a review see ref. 2), few studies have focused on elucidating the components of neural networks involved in pectoral motor control. These include direct and indirect top–down cortical projections to pectoral motoneurons (MN) in mice and primates (e.g., refs. 3–7) and local interneuron populations controlling reflexive actions in tetrapods (e.g., ref. 8). Studies on the salamander genus Necturus (9, 10), on mice (e.g., refs. 4, 6, 7, 11, and 12), freshwater hatchet fish (13, 14) and on zebrafish (e.g., refs. 15–17) have started to provide more information on how pectoral motor control is organized, yet the local pectoral network organization still remains enigmatic. Due to the MNs’ essential role as transmitters of motor commands to muscles, their anatomical and physiological characteristics play a pivotal role in controlling pectoral appendages. This importance is reflected in the organizational complexity of MN pools, which mirrors the complexity of the associated limbs. In tetrapods, motor pools controlling individual limb muscles are organized rostro-caudally, as well as along the medio-lateral extent of the spinal cord (SC), which allows for activation of distinct MN pools (18–22). Apart from a spatial organization of motor pools to control different pectoral muscle groups, an adequate organization of individual motor units formed within a given motor pool is needed to control muscles for varying behavioral amplitudes/speeds. To this date, it remains unclear, how pectoral motor pools are physiologically organized, beyond the size principle (23, 24). Teleost fishes offer several advantages to address pectoral MN organization as they possess two main muscle groups (here referred to generally as abductor [ABD] and adductor [ADD]) controlling their pectoral fins, while sharing an evolutionary ancestry with tetrapod pectorals (e.g., refs. 25 and 26). Similar to mammals, they exhibit a rostro-caudal separation of motor pools that innervate the fins through distinct nerves (27), but lack clearly separated, muscle-specific motor pools, at least in larval zebrafish (16, 17, 28). In most teleosts, pectoral fins are used for slow locomotion, maneuvering, and balance control (29). Due to the wide range of directions and amplitudes of the associated fin motions (for reviews see refs. 30–32), the activity of ABD and ADD likely share similar features with respect to activation strength, amplitude, and timing. This makes it difficult to investigate the organizational principle of pectoral motor pools underlying different behavioral amplitudes and velocities. Freshwater marbled hatchet fish (Carnegiella strigata) are a notable exception: They are well known for their hypertrophied pectoral ABD muscles mediating fast and large amplitude pectoral motions, which are used to propel the fish out of the water during escape behavior (13, 14, 33, 34). Like other fish, they also produce low-amplitude fin movements, presumably for stabilization and balance control. Thus, hatchet fish show vastly different degrees of pectoral motions between ABD and ADD, which need to be differentially controlled and are therefore ideal candidates to investigate adaptations to amplitude-dependent pectoral control. We used C. strigata to ask how additional neural flexibility is added to a pectoral network to increase its behavioral range. Using high-speed imaging and a series of anatomical and physiological experiments, we show that the motoneurons mediating pectoral activation are spatially organized into a mediolateral and dorsoventral map that differentiates the behavioral range of hatchet fish pectoral motions. The addition of larger-sized motoneurons in the ABD population with highly adapted electrophysiological properties provides an extension of the pool of MNs that control the production of a distinct behavioral motor pattern involving increased pectoral amplitudes and speeds. We propose that the addition of differently adapted motoneurons, organized in anatomical maps, is a general feature of motor pool adaptation, which has led to the diversity of pectoral motor control and thus motor behaviors in vertebrates in general. Results Behavioral Quantification. Freshwater hatchet fish are renowned for ballistic aerial excursions (jumping), as part of their flight reflex (33). Using high-speed video recordings to quantify pectoral fin motion amplitudes and velocities we were able to observe several aspects of pectoral fin motion during their jumping behaviors, that explain observations discussed in a previous study by Wiest (33): First, the pectoral fins are moved to the front, likely to fully open the fin surface, prior to the start of downward motion (Fig. 1 A, Upper and Movie S1). Second, we regularly observed a breaking of the water surface due to this forward motion (Movie S2). This could help to decrease the adhesion forces at the water–air interface. Third, the downward motion of the opened fins leads to two half-circle indentations (Movie S2) of the water surface, generating the previously described air pockets along the side of the fish and pushing the water along the keeled body (Movie S3–front view). This likely helps to reduce fish–water adhesion forces by leaving a smaller surface at the fish–water interface prior to becoming airborne. Fourth, the jump is not only generated by the pectoral fin abduction but also includes tail activation through the C-start reflex, as previously reported (33) and substantiated in our observations. Pectoral behavior of hatchet fish and behavioral quantification. (A) Visualization of pectoral fin motion during jumping (Top) and flickering (Bottom). Pectoral fins are color coded for better visualization. The final column shows superimposed shaded areas of pectoral fins from still images. The Inset displays a magnification of the circled area in the flickering schematic for easier visualization of fin tip difference during the flickering. (B) Color-coded comparison of pectoral fin abduction amplitude (fin tip position relative to the base of the fin) over time for both jumping (blue) and flickering (red) behavior. Mean is plotted in thick lines. Note that pectoral fin tips were not visible once fish left the water, reflected in the cut-off blue traces. (C–E) Boxplots for maximum amplitude (C), latency (D), and velocity (E) during fin abduction. Comparisons of amplitude, latency, and velocity of jumping and flickering were performed using the nonparametric, two-tailed Mann–Whitney U test (amplitude: P < 0.001, latency: P = 0.08, velocity: P < 0.001); for details, see Material and Methods. For jumping: N = 21, n = 33, for flickering N = 12, n = 43. [Scale bar in (A) represents 1 cm.] To evaluate the range of pectoral motion, we quantified fin motion during jumping, as well as during minute pectoral fin motions not involved in jumping behavior, referred to as flickering (Fig. 1 A, Lower and Movie S4). As the pectoral fins were moved exclusively downward during jumping behavior, we focused on downward flickering motions for better comparability. Compared to the jumping behavior (Fig. 1 B–E, blue), flickering motions (Fig. 1 B–E, red) showed 10.5 times smaller fin deflection amplitude (Fig. 1 B and C), 2.1 times longer latency to maximum fin deflection (Fig. 1D) and 10.3 times lower mean fin velocity (Fig. 1E; for details, see SI Appendix). Thus, our analysis revealed that hatchet fish are able to control the production of small as well as large pectoral amplitudes and velocities for distinct behavioral actions. Differential Muscle Volumes. Microcomputed tomography (µCT) was performed after tissue-contrast enhancement by means of staining a fixed C. strigata specimen in 3.5% B-Lugol’s solution. This iodine stain allowed to identify and segment pectoral muscles from reconstructed µCT images (Fig. 2). A previous report showed that the pectoral abductor and adductor are composed of several muscles (33). Here, we refer to these muscle groups as ABD and ADD muscles, respectively. µCT imaging and subsequent segmentation of the muscle groups in 3D revealed that the volume of pectoral ABD muscles was 27× larger than the ADD muscle volume (46.1 mm3 and 1.7 mm3, respectively). As previously described (33), hypertrophied ABD muscles were attached to the enlarged coracoids, which fused and formed a staggered disk dividing the animal’s keel throughout its vertical extent (Movie S5). µCT scan of a C. strigata individual with reconstructed ABD (blue) and ADD (red). Overview images in (A–C) show the fish skeleton in transparent gray for orientation. (A) Hatchet fish scan and reconstructed pectoral effector muscles from above. Left: overview, Right: detailed image in the same orientation. (B) Hatchet fish scan and reconstructed pectoral effector muscles from the front. Left: overview, Right: detail in the same orientation. (C) Hatchet fish scan and reconstructed pectoral effector muscles from the side. Top: overview, Bottom: detail in the same orientation. Scale bars: (A) (overview) = 2 mm, (A) (detail) = 1 mm; (B) (overview and detail) = 1 mm; (C) (overview) = 5 mm, (C) (detail) = 2 mm, Colors: red: ADD, blue: ABD, gray: fish. Abbreviations: ABD = abductor muscle, ADD = adductor muscle, N = 1. Pectoral Innervation. To understand how differential pectoral control is achieved in the hatchet fish, we identified the nerves innervating the pectoral fins. Sudan Black staining revealed three nerves (p1 to p3, SI Appendix, Fig. S1) innervating the fin muscles on each side, in line with the different segmental origin of the pectoral fin muscles. The most rostral nerve (p1) runs anterior of the pectoral fin base, while p2 and p3 run posterior to the fin base. Backfills of pectoral MNs via these nerves revealed a broad continuum of different MN sizes (measured via cross-sectional soma area) ranging from 44.5 to 2,551.9 µm2 (mean = 441.5 ± 428.42 µm2 (SD), n = 148 MNs, N = 8 fish; Fig. 3 A–D). Most neurons showed an elliptic to round shape with a mean ratio of minimum to maximum soma diameter of 0.64 ± 0.17 (SD, range: 0.24 to 0.99). Prominent large-diameter MNs with large axons (Fig. 3 A and D, blue arrowheads) were observed close to the central canal (CC). Pectoral motoneurons diverge in sizes. (A) Large backfilled motoneurons located close to the CC and (B) smaller MNs located more ventro-laterally (after backfills of nerve p2 and p3 respectively) extend their dendritic fields laterally. Asterisks indicate putative sensory input entering dorsally. (C) Histogram of maximum soma cross-sectional area; N = 8, n = 262, bin size = 75 µm2. (D) Confocal z-projection of a cleared hindbrain/spinal cord nerve backfill of p2. The blue arrow points to a very large motoneuron. The magenta asterisk marks putative sensory input entering the CNS dorsally. The double arrow indicates rostro-caudal orientation. The Inset in (D) (Top) depicts the location of panels (A and B). The Inset in (D) (Bottom) shows a partial 3D reconstruction of two individual MNs showing the rostro-caudal extent of MN dendrites. Scale bars are 50 µm for (A and B) and 100 µm for (D) and Bottom Inset in (D). Blue arrowheads depict exceptionally large MNs in (A and D). Abbreviations: c = caudal, CC = central canal, hb = hindbrain, ml = midline, p2 and p3 = pectoral nerve 2 and 3 respectively, r = rostral, sc = spinal cord. As for other teleost fishes (16, 27, 35), hatchet fish pectoral MNs were distributed within the caudal hindbrain/SC boundary extending into the first spinal segments. MN dendritic fields were oriented laterally (Fig. 3 A and B) and in a rostro-caudal direction (Fig. 3D) and did not cross the midline, which suggests that coupling of both sides is exclusively accomplished by common premotoneuronal innervation. One such premotoneuron is the giant fiber (13, 14), which mediates the jumping behavior. To assess the rostro-caudal extent of the processes, we partially 3D-reconstructed MNs (N = 2 fish, n = 18 neurons) in cleared whole-brain preparations (Fig. 3D and Movie S6). Again, differently sized neurons were apparent (Fig. 3D), which showed similar lateral dendritic arbor organization independent of their size and revealed a prominent rostro-caudal extent of the dendrites. MN axons exited the CNS ventrally (Fig. 3 A and B orange arrows). Additional projections from these tracings included the lateral line, the efferent nucleus, and putative sensory inputs (Fig. 3 A and B; D, magenta asterisks, SI Appendix, Fig. S2). To differentiate which MNs innervate the ABD and ADD muscles, respectively, we applied different tracers directly to the two muscle groups. As only the ABD underwent hypertrophy in the hatchet fish, the ADD was used as a reference for a more general teleost pectoral MN organization. We found a similar rostro-caudal distribution of ABD and ADD MNs along the pectoral MN column (Fig. 4 A and B). Despite the overlapping rostro-caudal distribution of MNs, a notable difference in MN sizes innervating the respective muscles was apparent: The range of ABD MN sizes included significantly larger neurons compared to the ADD (Fig. 4C, MN-ABD: Range: 33.4 to 2,270.5 µm2, mean ± SEM: 598.41 ± 42.07 µm2, n = 146; MN-ADD: Range: 26.9 to 1,080.2 µm2, mean ± SEM: 192.09 ± 18.66 µm2, n = 100; P < 0.001, Mann–Whitney U, N = 5 fish). These backfill experiments thus unambiguously attributed the above-mentioned large MNs to the ABD. These are likely an adaptation to the hypertrophied muscle used in the hatchet fish flight reflex. ABD and ADD motor columns are segregated along the mediolateral and dorsoventral axes. (A) Single spinal cord slice showing labeled neurons after tracer injections into the ABD (cyan) and ADD (magenta) muscle. (B) Rostro-caudal extent of the ABD (blue) and ADD (red) MNs for four individuals each, aligned to the most rostral nerve containing fibers backfilled through muscle injections. (C) Cumulative distribution functions of neuron cross-sectional area and histograms of neuron count of ABD (blue) and ADD (red) MNs. Note the shift of slope at ~300 µm2 that indicates an extension of the ABD MN pool toward bigger neuron sizes. (D–I) Distribution of motoneurons along relative x- and y-positions within the slice, collapsed over the entire motor columns (N = 5 fish). Marginal plots show boxplots and kernel density estimations (KDE) of the distributions. (D and E) show ADD MNs (n = 100), (F) shows ABD MNs (n = 146), (G) shows ABD MNs < 1,080 µm (n = 120), (H) shows all MNs (n = 246) and (I) shows all MNs < 1,080 µm (n = 220). [Scale bar in (A) is 100 µm.] Abbreviations: ABD = abductor, ADD = adductor, MNs = motoneurons. Spatial Segregation of ABD and ADD MN Pools. The medio-lateral and dorso-ventral distribution of ABD (N = 5 fish, n = 146 neurons) and ADD (N = 5, n = 100) MNs showed an overlapping, yet significant (P < 0.001, Mann–Whitney U) spatial separation into two motor columns across their rostro-caudal extent (Fig. 4 D–I): ABD MNs were located more medially and more dorsally (Fig. 4A, cyan; Fig. 4 F and H, blue; mean relative medio-lateral ABD MN-position: 0.39 ± 0.008; mean relative dorso-ventral ABD MN-position: 0.28 ± 0.007), while ADD MNs were distributed more laterally and more ventrally (Fig. 4A, magenta; Fig. 4 D, E, H, and I, red; mean relative medio-lateral ADD position: 0.50 ± 0.008; mean relative dorso-ventral ADD position: 0.21 ± 0.008). To test whether this differential distribution depended on the large ABD MNs, absent in the ADD, we performed a second analysis in which we removed the large ABD neurons (Fig. 4 G and I). Large ABD neurons were defined as those that exceeded the maximum size of those measured in the ADD tracing experiments (max. soma area in ADD was 1,080 µm2). Again, two distinct, yet overlapping MN pools were present (mean relative medio-lateral position of small ABD MNs: 0.41 ± 0.007, n = 120; mean relative medio-lateral position of ADD MNs: 0.50 ± 0.008, n = 100, P < 0.001, Mann–Whitney U; mean relative dorso-ventral position of small ABD MNs: 0.26 ± 0.007, n = 120; mean relative dorso-ventral position of ADD MNs: 0.21 ± 0.008, n = 100, P < 0.001, Mann–Whitney U). Our data thus show that, despite an overlap of the two different MN pools, a segregation into different pools is present across the medio-lateral and dorso-ventral extent of the SC. Hence, although less elaborate, the organization of pectoral motor pools innervating distinct muscle groups in the freshwater hatchet fish is reminiscent of that seen in tetrapods. Size-Dependent Organization of Motoneuron Pools. In addition to the muscle-specific distribution of pectoral MNs, backfills further revealed size-dependent medio-lateral and dorso-ventral distributions of MNs: Larger MNs were located more dorso-medially, while smaller MNs were found more ventro-laterally (Fig. 5 A and B; see SI Appendix, Fig. S3 for statistics) in the ABD MN pool (P < 0.001 and P < 0.001, respectively). While there was a trend toward a similar, size-dependent, medio-lateral and dorso-ventral separation in the ADD MN pool, it was not significant (P = 0.176 and 0.084, respectively; see SI Appendix, Fig. S3 for statistics). Larger MNs are expected to form bigger motor units that contribute to simultaneous muscle activation (13, 14). Thus, we predict that the large MNs, which were backfilled by tracer application to the ABD, contribute to the large amplitude and high velocity behavior associated with aerial escape. As both ABD and ADD contribute to the flickering behavior, both corresponding MN populations also share similarly sized, smaller MNs. Size-dependent distribution of pectoral MNs. (A) Labeled pectoral MNs after tracer injections into ABD and ADD muscles. (B) Larger (dark orange and purple) ABD MNs are located more dorsomedially than smaller ones (N = 5, n = 146). (C) No mediolateral or dorsoventral size dependency could be found across ADD MNs (N = 5, n = 100). (D) Larger MNs are generally located more dorsomedially, when accounting for ABD and ADD respectively. [Scale bar in (A) is 100 µm.] Abbreviations: ABD = abductor, ADD = adductor, CC = central canal, MNs = motoneurons. Electrophysiological Properties of Pectoral Motoneurons. Due to the difference in organization of MN sizes between ABD and ADD MN pools, we next tested whether the electrophysiological properties of MNs varied depending on their size. We performed whole-cell patch clamp recordings of MNs of different sizes (cross-sectional areas between 110 µm2 and 2,041 µm2 with a mean of 569.22 ± 133.4 µm2; N = 11 fish, n = 16 neurons, Fig. 6A). These MN sizes corresponded to the range of areas measured in our backfill experiments (Figs. 3 and 4). We used low-amplitude, negative current pulses to probe membrane properties close to resting membrane potential of patched neurons (Fig. 6 B and C). The input resistance varied with soma size and showed a significant trend to decrease with increasing soma size (Fig. 6B), as expected according to the size principle (24). The average time constant Tau of patched neurons was 7.1 ms ± 1.9 ms (n = 16, Fig. 6C). The average rheobase was 416.38 ± 132.6 pA (n = 16, Fig. 6D). AP firing threshold voltages of patched neurons lay at −46.67 ± 0.95 mV and did not show differences that depended on soma size (Table 1). Action potential (AP) properties, including amplitudes of AP and AHP (after-hyperpolarization), as well as their full width at half maximum height (FWHM) were not correlated to soma size (Table 1). Electrophysiological properties of pectoral motoneurons of the hatchet fish. MNs can be separated into two groups, according to their firing properties (“integrator,” MN-I, orange and “onsetter,” MN-O, petrol). (A) Soma sizes of patched pectoral MNs corresponded with those measured in backfill experiments (range: 110 to 2,041 µm2, 569.22 ± 133.4 µm2). MN-I displayed significantly smaller soma areas than MN-O. (B and C) Basal membrane properties of MN-I and MN-O differed significantly, with MN-I having higher input resistance (B) and slower time constants (C). Rheobase currents were significantly higher in MN-O (D). (E–K) Stimulation with rheobase elicited action potentials with significantly longer onset latencies (E, F, and I) and broader jitter (E, F, and J) in MN-I. The strong onset properties of MN-O were caused by a pronounced onset-repolarization which became visible when stimulation at rheobase did not trigger AP firing and was not present in MN-I (G, H, and K). Stimulus strength: 610 pA in (E) and 95 pA in (F). *P < 0.05, ***P < 0.001. For exact values and statistics, see Table 1. Table. 1.Summary of physiological parameters of hatchet fish pectoral motor neuronsunitAll MNs (N = 11, n = 16)MN-O (N = 7, n = 8)MN-I (N = 6, n = 8)P-valueSoma cross-sectional area[µm2] 569.22 ± 133.4 (110 to 2,041) 914.29 ± 203.14 (330.4 to 2,041.02) 224.14 ± 31.24 (109.8 to 327.14) 1.55*10−4 Input resistance[MΩ] 164.62 ± 55.66 (5.43 to 832.82) 53.79 ± 22.96 (5.43 to 161.77) 275.45 ± 96.13 (67.40 to 832.82) 0.0148 Tau[ms] 7.08 ± 1.92 (1.07 to 28.35) 3.07 ± 0.81 (1.07 to 4.73) 11.10 ± 3.24 (3.54 to 28.35) 0.0148 Rheobase current[pA] 416.38 ± 132.60 (6.00 to 1,763.00) 739.50 ± 211.45 (57.00 to 1,763.00) 93.25 ± 28.41 (6.00 to 260.00) 0.0281 AP threshold[mV] −46.67 ± 0.95 (−53.33 to −40.23) −47.24 ± 1.42 (−51.39 to −41.98) −46.10 ± 1.32 (−53.33 to −42.53) 0.5054Onset integral[s*V*10−5] −10.57 ± 7.37 (−95.30 to 20.77) 10.78 ± 2.43 (1.75 to 20.77) −31.93 ± 9.83 (−95.30 to −12.13) 1.55*10−4 Onset latency[ms] 291.73 ± 79.16 (8.94 to 815.12) 23.01 ± 6.28 (8.94 to 63.61) 560.44 ± 78.64 (815.12 to 172.92) 1.55*10−4 Jitter[ms] 103.87 ± 28.31 (1.45 to 316.03) 9.51 ± 6.07 (1.45 to 51.63) 198.23 ± 29.21 (64.31 to 316.03) 0.0002 Maximum mean AP frequency[Hz] 261.87 ± 36.51 (37.85 to 514.69) 181.33 ± 48.14 (37.85 to 414.69) 342.41 ± 39.27 (175.41 to 514.69) 0.0281 AP-amp[mV] 95.66 ± 2.68 (81.84 to 119.37) 93.78 ± 3.98 (81.84 to 100.79) 97.54 ± 3.72 (83.80 to 119.37) 0.4418AHP-amp[mV] −19.19 ± 1.10 (−26.83 to 12.36) −18.70 ± 1.34 (−24.92 to 13.87) −19.68 ± 1.83 (−26.83 to -12.36) 0.6454AP-FWHM[µs] 143 ± 9 (110 to 250) 149 ± 16 (110 to 250) 136 ± 8 (110 to 180) 0.6946AHP-FWHM[ms] 1.02 ± 0.30 (0.37 to 5.29) 0.81 ± 0.14 (0.40 to 1.48) 1.23 ± 0.60 (0.37 to 5.29) 0.4580Upper entry is mean ± SEM. Lower entry gives ranges of respective parameters. All significance tests performed with the nonparametric Mann–Whitney U test. Summary of physiological parameters of hatchet fish pectoral motor neurons 569.22 ± 133.4 (110 to 2,041) 914.29 ± 203.14 (330.4 to 2,041.02) 224.14 ± 31.24 (109.8 to 327.14) 164.62 ± 55.66 (5.43 to 832.82) 53.79 ± 22.96 (5.43 to 161.77) 275.45 ± 96.13 (67.40 to 832.82) 7.08 ± 1.92 (1.07 to 28.35) 3.07 ± 0.81 (1.07 to 4.73) 11.10 ± 3.24 (3.54 to 28.35) 416.38 ± 132.60 (6.00 to 1,763.00) 739.50 ± 211.45 (57.00 to 1,763.00) 93.25 ± 28.41 (6.00 to 260.00) −46.67 ± 0.95 (−53.33 to −40.23) −47.24 ± 1.42 (−51.39 to −41.98) −46.10 ± 1.32 (−53.33 to −42.53) −10.57 ± 7.37 (−95.30 to 20.77) 10.78 ± 2.43 (1.75 to 20.77) −31.93 ± 9.83 (−95.30 to −12.13) 291.73 ± 79.16 (8.94 to 815.12) 23.01 ± 6.28 (8.94 to 63.61) 560.44 ± 78.64 (815.12 to 172.92) 103.87 ± 28.31 (1.45 to 316.03) 9.51 ± 6.07 (1.45 to 51.63) 198.23 ± 29.21 (64.31 to 316.03) 261.87 ± 36.51 (37.85 to 514.69) 181.33 ± 48.14 (37.85 to 414.69) 342.41 ± 39.27 (175.41 to 514.69) 95.66 ± 2.68 (81.84 to 119.37) 93.78 ± 3.98 (81.84 to 100.79) 97.54 ± 3.72 (83.80 to 119.37) −19.19 ± 1.10 (−26.83 to 12.36) −18.70 ± 1.34 (−24.92 to 13.87) −19.68 ± 1.83 (−26.83 to -12.36) 143 ± 9 (110 to 250) 149 ± 16 (110 to 250) 136 ± 8 (110 to 180) 1.02 ± 0.30 (0.37 to 5.29) 0.81 ± 0.14 (0.40 to 1.48) 1.23 ± 0.60 (0.37 to 5.29) Upper entry is mean ± SEM. Lower entry gives ranges of respective parameters. All significance tests performed with the nonparametric Mann–Whitney U test. At rheobase, we observed two categories of firing patterns: onset-firing with short latency vs. tonic firing with variable latency (Fig. 6 E, F, and I). We grouped neurons according to these firing patterns into two categories: “integrators” (MN-I, orange in Fig. 6) and “onsetters” (MN-O, petrol in Fig. 6). Soma size highly correlated with these two categories, as MN-O were larger than MN-I (Fig. 6 A–D and Table 1). MN-I displayed higher input resistance (Fig. 6B and Table 1) as well as longer time constants (Fig. 6C and Table 1) and fired APs at significantly lower rheobase currents (Fig. 6D and Table 1). Upon repeated stimulation at rheobase, MN-I showed significantly higher variability (“jitter”) in first AP latencies, compared to MN-O (Fig. 6 E, F, and J). These differences in jitter were due to a pronounced repolarizing onset current that appeared shortly after stimulus onset in MN-O but not MN-I (Fig. 6 G, H, and K). Discussion Freshwater hatchet fish, like many other fish, show slow, small-amplitude pectoral movements that are likely used for maneuvering/balancing. In addition, they produce unusual high-velocity, large-amplitude pectoral motions for their aerial excursions. Accordingly, they need to coordinate a wide range of different pectoral motion amplitudes and velocities, some of which are associated with distinct groups of pectoral muscles. Hatchet fish pectorals are therefore reminiscent to those of other vertebrates, including tetrapods, that need to control different muscles of more complex pectoral appendages to achieve effective control of pectoral movements in different amplitude and velocity ranges. We demonstrate that the pectoral motoneuron pools in marbled hatchet fish follow several distinct organizational principles, explaining their ability to achieve such differential pectoral control: i) As in other teleosts (15, 17, 27, 36), and in vertebrates in general (20–22), hatchet fish pectoral MNs are organized in rostro-caudal columns. Unique to most fish and in contrast to tetrapods, pectoral MN pools are located in the caudal hindbrain and rostral spinal cord (27). ii) In contrast to other fishes in which pectoral pools have been investigated [e.g., larval zebrafish and sea robins (17, 32)], the hatchet fish ABD is innervated by an additional pool of larger-sized MNs which is not present in the ADD. This MN pool correlates with the wider range of amplitudes and speeds that the ABD muscles need to produce, which explains its absence in the ADD. To our knowledge, similarly large pectoral motoneurons have not been reported in other fishes (e.g., refs. 15, 17, and 35), except for the butterfly fish Pantodon buchholzi, which has a similar pectoral escape behavior (37). This suggests that the addition of differentially sized MNs innervating the ABD reflects an evolutionary adaptation of pectoral motor networks to accommodate for pectoral behaviors of wide-ranging amplitudes and speeds. iii) A third organizational principle is a medio-lateral and dorso-ventral separation of motoneurons innervating different muscle groups, here ABD and ADD. While an overlapping population of MNs innervating the ABD and ADD might be advantageous to generate muscle tone and keep the fin stable, differential activation of both ABD and ADD is needed. The spatial separation of MN pools might contribute to such differential activation by segregating synaptic inputs. Such a topographic representation of pectoral muscle control is thus reminiscent to that seen in tetrapods, which have evolved more complex motor behaviors with multijoint limbs, with even clearer motor pool separation. There, MNs form distinct and differently sized motor pools for individual muscles, which are spatially separated across both the rostro-caudal and medio-lateral extent of the SC (18, 20, 21, 38). While a spatial separation of motor pools controlling axial muscles for different swimming speeds is known from zebrafish (see next section), the medio-lateral and dorso-ventral separation of ADD and ABD MNs presented here has not been reported in pectorals of larval zebrafish (15, 17). This might be due to the smaller number of pectoral MNs in larvae or to the different range of pectoral motion amplitudes and speeds in zebrafish. The (partial) overlap of ADD and ABD MN pools in both hatchet fish and zebrafish raises the interesting question of how a coordinated activation of intermingled motoneurons by different premotoneurons is further organized beyond the principles of Hebbian innervation. iv) Hatchet fish exhibited a size-dependent, medio-lateral and dorso-ventral distribution of pectoral MNs across the pooled ABD and ADD. Intriguingly, such a separation has also been demonstrated for MNs innervating anatomically separated axial muscle groups governing different swimming speeds in zebrafish (39–41). There, large MNs are also located more medially in the SC and innervate “fast” musculature, while “intermediate” and “slow” muscles are innervated by MNs of decreasing size that are located more and more laterally. This size-specific organization likely provides a neural substrate map for motor behaviors in different behavioral ranges, as it might facilitate premotor synaptic input to different MN pools. Such size-dependent organization might thus be a general principle of spinal motor systems to organize MNs for different behaviors. To validate this hypothesis, premotoneuronal inputs will have to be mapped. While in hatchet fish the premotoneurons initiating the pectoral aerial escape behavior, the so-called giant fibers, are known (13, 14), other premotoneuronal populations that mediate MN activation for different pectoral motor behaviors (i.e., flickering) remain elusive. v) The size-related separation of MNs was also highly correlated with differently tuned physiological responses, generating a map of MN recruitment depending on both motoneuron size and physiology. As large motoneurons, that are located dorso-medially, were onsetter neurons (MN-O), they are ideally suited for synchronous muscle fiber activation needed to mediate the rapid pectoral flight behavior. The more ventro-laterally located motoneurons were smaller and of the integrating type (MN-I). The correlation between varying neuron size and AP firing characteristics was also reflected in other physiological parameters such as input resistance, rheobase currents, or firing precision that all depended on soma size (Fig. 6), in line with the size principle and the organization of axial locomotor motoneurons in zebrafish (39–41). In our slice preparations, we were not able to identify whether the recorded MNs belonged to the ABD or ADD pools. However, the demonstrated MN size map is associated with different physiological attributes necessary for the production of different behavioral motor patterns: Escape responses should be elicited exclusively when needed, so a key control factor is preventing neurons from accidental firing. According to the size principle, excitability of neurons decreases with increasing MN size (23, 24, 41). As escape behaviors need to be fast and precise, large MNs benefit from phasic firing properties. Another characteristic of such large motoneurons is the fast conduction velocity of their axons, which also contributes to the suitability for mediating escape responses. Smaller neurons, in contrast, are ideally suited to perform low-amplitude flickering behavior for fine balance or steering control, which does not rely on short latencies and large abduction amplitudes. Such fine motor behavior should be activated in a graded fashion to be able to optimally align the body position to varying turbulence in water currents. Thus, the physiological properties of the different MN pools (that are defined by size, medio-lateral and dorso-ventral location) are highly adapted to elicit different behaviors. However, this physiological categorization does not exclude that MN-Is contribute to the jumping behavior. Rather, MNs might be recruited differentially during various phases of the jumping behavior. The interplay of these five organizational principles contributes to the execution of vastly different behaviors, in addition to premotoneuronal control, as it allows to differentially activate motoneurons that innervate different muscles in conjunction or independently. This results in pectoral motions of different kinematical profiles. A Deep Homology of MN Organization between Actinopterygian and Sarcopterygian Lineages? Due to the homology of pectoral fins and tetrapod limbs and the reduced biomechanical complexity (i.e., fewer muscle groups and joints) of pectoral fins in fish compared to tetrapod limbs, fish provide excellent opportunities to investigate the fundamental organization and evolution of pectoral networks. As either (fish or tetrapods) face similar control challenges, the organizational principles of the respective neuronal networks might share similar solutions allowing to better understand pectoral motor control in general. Among the different problems faced is the need to control limb-/fin-specific muscles simultaneously and independently of each other for muscle tone and muscle movements. The ADD and ABD motor pools of hatchet fish offer solutions for muscle control over vastly different behavioral ranges. Similarly, tetrapods require fine-scale control, especially relevant for complex limbs that have individually accessible muscles and joints associated with digits, as well as large and strong behavioral actions like those executed by the biceps, for instance. The orchestrated control of such different muscle strengths/motion amplitudes has been solved in tetrapods by spatial and/or physiological separation of motor pools (18–22, 24). The above-described organization of the hatchet fish pectoral motor pools, allowing the execution of different behaviors is in line with the one previously reported for tetrapods. While the size-dependent medio-lateral and dorso-ventral organization of pectoral MNs in hatchet fish has not been reported for tetrapod MNs, their motor pools are similarly made up of differently sized MNs with different electrophysiological profiles. Together this suggests a deep homology in vertebrate motor systems to organize MNs that innervate different effector muscles controlling motor behaviors. Summary. We show that hatchet fish share a set of organizational principles of their MN pools, that are reminiscent of those of tetrapods (18–22) and those known from axial muscle control in zebrafish (39, 41). This organization allows for controlling different pectoral motor behaviors with widely diverging amplitudes and velocities. These spatial and physiological organization principles provide insight into the mechanisms by which motor pools associated with different limb muscles and/or more complex appendage-related motor behaviors could have evolved from the existing, axial central pattern generators. In contrast to MNs innervating different axial muscle groups which allow swimming at different speeds, we show that a set of similar organization principles, especially a muscle-specific medio-lateral, as well as dorso-ventral organization of MNs, allows for an increased range of pectoral motor control. Materials and Methods Animals, Diet, and Maintenance. Twenty-three marble hatchet fish, C. strigata (Günther, 1864), of undetermined sex (average standard length: 2.9 ± 0.3 cm; average weight: 0.6 ± 0.2 g), were used for neuroanatomical experiments and behavioral observations. Animals were obtained as adults from a local pet shop (Zoo Muser, Graz, Austria) or from a pet fish supplier (Tropicwater, Rodgau, Germany, www.tropicwater.eu) and were held in groups of up to 30 individuals in glass aquaria (120 × 40 × 40 cm) at 25 °C. Upon arrival of new fish, they were granted an accommodation period of a week, before they were considered eligible for experiments. The fish were kept under a 13:11 light:dark rhythm and were fed daily with commercial dry food. All experiments were conducted under the guidelines of the Austrian Bundesministerium für Bildung, Wissenschaft und Forschung (protocol number: 285175223). Behavioral Observations. Video data were obtained at 539 fps (n = 16 jumping events, n = 43 flickering events) and 2,000 fps (n = 17 jumping events) using two high-speed video camera systems [539 fps: DMK 37AUX287 camera (TheImagingSource, Bremen, Germany); 2000 fps: LaVision High Speed Star 4, LaVision GmbH, Göttingen, Germany]. Images were recorded using IC Capture 24 (TheImagingSource) and Photron FASTCAM Viewer (v. 3.282, Photron, San Diego, CA), respectively. For the duration of the observations, 2 to 5 animals were moved to one of two experimental tanks at a time (60 × 35 × 30 cm, 20 × 25 × 20 cm), where a short vibrational impulse (a weight of 200 g was dropped onto the table carrying the aquarium from a height of 50 cm, which was not observable by the fish) was used to elicit jumping behavior. For flickering behavior, films were taken during resting periods of the fish. Videos were saved and motion tracking was performed on one fin at a time using a custom-written software (courtesy of Michael Hofmann, University of Bonn). For data analysis, a custom-made macro within IGOR PRO (v. 8.04, Wavemetrics, Portland, OR, United States) and a custom-written Python script (v. 3.8, Python Software Foundation, www.python.org), building on the NumPy [v.1.18.5, (42)], matplotlib [v.3.2.2, (43)], pandas [v.2.2.1, (44)] and seaborn [0.12.2, (45)] libraries were used. Statistics were performed using the scipy library [1.13.1, (46)]. µCT Scanning and 3D Tissue Reconstruction. For µCT imaging, one animal (standard length = 3.0 cm, body mass = 0.5 g) was killed by immersion in 0.05% MS-222 (PharmaQ Ltd, Overhalla, Norway), dissolved in tank water. The animal was subsequently fixated in 4% paraformaldehyde (PFA) in 0.1 M PB for 24 h at 4 °C. The animal was submersed in 3.5% B-Lugol’s iodine solution for 48 h at 4 °C to enhance contrast. B-Lugol’s was prepared by mixing a two-times Sorensen buffer (Na2HPO4: 266 mM, KH2PO4: 266 mM, pH: 7.2) with 7% aqueous Lugol’s solution (I2: 70 mg/mL, KI: 140 mg/mL) at a 1:1 ratio. B-Lugol’s was used to minimize tissue acidification and associated tissue shrinkage (47). After staining, the animal was rinsed with 0.1 M PB and embedded in 1% low-melt agarose (Agarose Low-Melt, Carl Roth GmbH+Co. KG, Karlsruhe, Germany) in 0.1 M PB for µCT imaging. Whole-fish μCT scans (one with the unstained and one using the stained fish) were performed using a Scanco µCT40 scanner (Scanco Medical AG, Brüttisellen, Switzerland) with the following scan settings: 70 kV, 114 µA, 8 W, 300 ms integration time, 3 averages, and 1,000 projections/180°. Custom-written Matlab code (R2023, The MathWorks, Inc., Natick, Massachusetts, USA) was used to extract reconstructed 16bit tiff image files from the original Scanco data files. Images were of dimensions 2,048 × 2,048 pixels with 10 µm resolution per pixel. Slice thickness was 10 µm, resulting in an overall voxel resolution of 10 µm. Reconstruction of ABD and ADD muscles from µCT data was performed manually using Amira (V. 2020.2, Thermo Fisher Scientific, Waltham, MA, USA). Quantification of Fin Movements. The position of the fin tip was evaluated in relation to the fin base: When both were located on the same horizontal plane, the relative fin tip position was zero. If the fin tip was located above (below) this line, the positional value was positive (negative). To compensate for passive drift and motion of the entire fish within the water, the fin tip position was measured relative to a pigment spot on the respective fin base. For both flickering and jumping events, a 1 mm fin deflection from resting position was chosen as the onset of a fin abduction event, which filtered out passive fin deflections induced by water currents in the tank. Abduction peak amplitude was calculated by measuring the distance between resting position (normalized to zero) and the peak amplitudes. Abduction peak latency was then calculated as the time between the onset and the deflection peak. Since the deflection peak for both jumping and flickering behavior was situated below the water surface, we refrained from above-water image acquisition and analysis. Thus, observations of flight trajectory and its ballistic nature were not conducted, the reader is kindly referred to the work of Wiest (33) for more details on hatchet fish escape behavior. Velocity was calculated as the first derivative of the amplitude values for each behavior, respectively. Unless stated otherwise, means are always presented with the SEM. Whole-Fish Staining with Sudan Black. Nerves innervating the pectoral fins (hereafter referred to as pectoral nerves; NB: These nerves do not exclusively innervate the pectorals; see results) were visualized by means of Sudan Black staining, using a modified version of the protocol by Song and Parenti (48). Fish (N = 3) were killed by immersion in 0.05% MS-222 (PharmaQ Ltd, Overhalla, Norway) in tank water for 15 min and subsequently fixed with 4% PFA in 0.1 M PB overnight. Before staining, animals were placed in 0.1 M PB overnight to remove the fixation solution. The following steps were conducted at room temperature, unless explicitly stated otherwise. Depigmentation was conducted with a solution of 1.5 mL 30% H2O2 in 100 mL 0.5% KOH for 24 h exposed to sunlight. Next, specimens were treated with a maceration solution of 2 g trypsin from porcine pancreas (Sigma-Aldrich, St. Louis, MO) in 200 mL 0.1 M PB and remained in this solution for 24 to 36 h until they were almost completely translucent. Consecutively, animals were submersed in 30%, 50%, and 70% ethanol for 30 min, respectively. For the staining solution, 0.5 g of Sudan Black B (Carl Roth) was dissolved in 100 mL 70% ethanol in a warm-water bath at 40 °C. The saturated solution was filtered and diluted with 70% ethanol to obtain a 30% Sudan Black B working-solution. Animals were left in this solution overnight. Destaining was initialized through a differentiation period, in which each animal was dipped into 70% ethanol for 30 s, then into 60% ethanol for 1 min, and finally placed into 50% ethanol on a shaker at 150 rpm until the tissue proved near-translucent again. Mechanical acceleration of the process could be induced by squeezing excess staining solution gently out of the animal with forceps. When an animal had reached the desired state of translucency, it was put into 0.5% KOH overnight. The next day, the superficial skin layer was removed for better visualization of the pectoral innervation. For storage, specimens were transferred to 40% glycerol in 0.5% KOH for 3 h, followed by an overnight immersion in 70% glycerol in 0.5% KOH. Finally, stained fish were placed into 100% glycerol for storage. The stained animals were imaged using a binocular (Nikon SMZ 25, Minato City, Tokyo, Japan) and images were acquired using the software NIS-Elements BR 5.20.00 (Nikon). Motoneuron Backfills. Individual fish (N = 11) were anesthetized with a 0.025% MS-222 solution dissolved in tank water. Scales were removed from the surgical spot and two superficial incisions into the skin were made to expose individual nerves innervating the pectoral fin muscles. A small piece of parafilm was slid under the target nerve. The most anterior of the three nerves was more difficult to access. Therefore, the overlying muscle tissue was removed before the nerve was cut and tracer applied using tracer-coated insect needles. Depending on the experiment, different tracers and tracer-combinations were applied (i.e., Dx-Alexa 488, 10 kDa; Dx-Alexa 546, 10 kDa; BDA, 3 kDa with or without neurobiotin; Invitrogen by ThermoFisher, Eugene, OR). In cases where solely neurobiotin was applied, small insect needles were used to transfer the neurobiotin crystals directly onto the nerve stump. After 5 min, excess dye was suctioned with a tissue wipe (KIMTECH wipe, Kimberly Clark Professional; Roswell, GA). Subsequently, a small piece of tissue wipe, soaked with 0.25% Bupivacaine (Jenapharm, Jena, Germany) for local anesthesia, was applied as wound patch, before the wound was closed using tissue adhesive (3 M Vetbond, 3 M Animal Care Products, St. Paul, MN). After surgery, fish were transferred into a beaker with tank water and left to recover for at least 30 min in a dark container before they were transferred back to the holding facility for an overnight incubation period. The next day, the fish were deeply anesthetized with 0.05% MS-222 in tank water for 15 min and subsequently decapitated. Brains were exposed and fixed in 4% PFA in 0.1 M PB overnight. After fixation, brains and the rostral part of the spinal cords were extracted, embedded in 4% agarose (Carl Roth) in 0.1 M PB and sectioned to 100 µm transversal slices on a vibratome (Campden Instruments, 7000smz-2, Loughborough, UK). In cases where BDA or neurobiotin were used as tracer molecules, slices were left to dry overnight before they were washed twice in 0.1 M PB and subsequently incubated in 0.1 M PB containing fluorescently tagged streptavidin (dilution 1:500; Life Technologies, Carlsbad, CA). Sections were then observed under a fluorescent Microscope (Axioplan, ZEISS, Oberkochen, Germany) and selected slices were imaged with confocal microscopes (SP8, Leica, Wetzlar, Germany, and Cerna Confocal CM712, Thorlabs, Newton, NJ). Image processing and anatomical quantification were performed in Fiji [v. 1.52, (49)]. The total number of labeled neurons was determined, their major and minor soma diameters, as well as the cross-sectional area measured and the location of each neuron within the respective sections determined. To compensate for brain and fish sizes, the position of each cell, as defined by the center midpoint of their soma (n = 158 neurons, N = 8 animals) was taken as their x- and y- positional values within the slice. Then, they were normalized to the size of the slice. We used a custom-written Python script to plot the distribution of the cell sizes. For this, we used the libraries pandas (44) and seaborn (45). For the size histogram, we chose a 75 µm² bin size. In two animals, brains were not sectioned after tracing. Instead, brains were cleared after extraction, following the protocol of Kuwajima et al. (50). Cleared whole-brains were imaged with a confocal microscope (Leica SP8). Images were stitched and stacked (Fiji) for 3D reconstruction (Neurolucida 360, v2017.01.3, MBF Bioscience, Williston, VT). Tracer Injections into the Muscle. In 5 animals (standard length: 2.3 ± 0.2 mm; body mass: 0.3 ± 0.0 g; mean ± SEM), we performed tracer injections into the ABD and ADD muscle groups. For this, animals were prepared as described above, but instead of labeling the nerve, small incisions were made on the adductor muscle of the pectoral fin, and Dx-tetramethylrhodamine-crystals (3 kDa, Invitrogen by Thermo Fisher Scientific), were inserted into the incision site. In the same animal, we added a small incision on the abductor muscle and inserted small crystals of neurobiotin into the incision. The rest of the procedure was as listed for nerve backfills above. In two of these muscle injection experiments, we performed 100 µm sections on the vibratome. For the other animals, we cryoprotected the tissue overnight in 30% sucrose in 0.1 M PB and sectioned the tissue to 50 µm in the transverse plane on a cryostat (CM3000, Leica and SLEE MEV, SLEE medical GmbH, Nieder-Olm, Germany). In Vitro Slice Patch Clamp Electrophysiology. For in vitro slice patch clamp electrophysiology, animals (N = 11) were deeply anesthetized in 0.05% MS-222 for at least 15 min. Extraction of the brain was performed in ice-cold, oxygenated slicing solution (in mM: sucrose: 50, NaCl: 25, NaHCO3: 27, KCl: 2.5, MgCl2: 3, CaCl2: 0.1, glucose: 25, ascorbic acid: 0.4, myoinositol: 3, Na-pyruvate: 2) at 4 °C. After extraction, brains were embedded in low-melt agarose (2 to 4%, Agarose Low-Melt, Carl Roth) and cut to 250 µm sections in the transverse plane using a vibratome (Ci 7000 SMZ 2, Campden Instruments). Sections were left to recover for at least 30 min in recording solution (in mM: same as slice solution but with 125 mM NaCl, 2 mM CaCl2, 1 mM MgCl2, and no sucrose). Slices were then transferred to a recording chamber and continuously superfused with patch clamp recording solution (see above) at room temperature. A microscope (Zeiss Examiner.A1) with Dodt contrast and a 40× immersion objective (Zeiss 40×/1.0 DIC VIS-IR) was used to visualize the motor neurons. A camera (Orca flash 4.0 LT, C11440, Hamamatsu Photonics K.K., Hamamatsu, Japan) was used to show slices on a computer monitor and target motor neurons with glass micropipettes for current clamp recordings. A micropipette puller (P-87 Micropipette Puller, Sutter Instrument, Novato, CA) was used to pull patch clamp electrodes. For recordings, electrodes were filled with patch solution (in mM: K-gluconate: 145, KCl: 4.5, HEPES: 15, Mg-ATP: 2, K2-ATP dihydrate: 2, Na2-GTP: 0.3, Na2-phosphocreatine: 7.5) to reach a final resistance of 4 to 7 MΩ. The liquid junction potential was not corrected. Motoneurons were targeted through a visual approach: We only used hindbrain/spinal cord sections containing the pectoral motor column, as previously identified in our tracing experiments. We targeted neurons in a diagonal axis, from the CC to the lower corner of the tissue section, which we identified as the main axis of the pectoral motor column through our tracing experiments. Neurons were passively filled with dye contained in the patch pipette during recordings, to visualize the neurons’ morphology. Stimulus Protocols. Basal membrane properties. To measure basal membrane properties, 40 repetitions of 500 ms current pulses of −5 to −50 pA were applied to patched neurons at resting membrane potential with no holding current. The mean resulting potential drop was used to calculate input resistance (Rin), and the time constant tau (τ). Rheobase. To measure rheobase currents needed to bring neurons to their AP firing threshold, square current pulses of 1,000 ms length and consecutively increasing amplitudes were applied. First, applied current amplitudes were increased stepwise in 25 µA increments to get a rough estimation of rheobase current. For exact rheobase measurement, the same protocol was used. However, current amplitude was increased by 1 pA between individual current steps, starting with a stimulus amplitude estimated to lie slightly below rheobase. Action potential properties. To measure the properties of individually elicited APs, ramp stimuli with a rise time of 180 µs and a decay time of 1 ms were applied to mimic excitatory postsynaptic currents. The amplitude of the applied stimulus was increased in 100 pA steps. AP analysis was performed after subtracting the last subthreshold recording from the first suprathreshold recording, to adjust for subthreshold potential-changes and allow for a direct measurement of only the AP defining parameters. Measurements included the amplitude of the AP and the afterhyperpolarization (AHP), as well as the FWHM of AP and AHP. AP firing precision. To measure the precision of hatchet fish pectoral motor neuron firing, we applied twenty repetitions of 1,000 ms current pulses at rheobase. The mean and SD of spike occurrence were calculated as a measure for the latency and the “jitter” in a neuron’s response. To quantify and visualize differences in the subthreshold voltage responses of MNs, the integral between recording traces and a fictive horizontal line at the height of the new equilibrium potential was calculated for nonspiking repetitions. Statistics. Behavioral analysis. Data for maximal amplitude, latency, and velocity for both jumping and flickering were tested for normality using the Shapiro–Wilk test. Maximal amplitude, latency, and velocity were normally distributed for jumping behavior (Shapiro–Wilk, amplitude: P = 0.7807; latency: P = 0.0847 and velocity: P = 0.5793, N = 21, n = 33) but not during flickering behavior (Shapiro–Wilk, amplitude: P < 0.01; latency: P < 0.001 and velocity: P < 0.001, N = 12, n = 43). Therefore, further comparisons were conducted with nonparametric tests. For comparison of maximum amplitudes, velocities, and latencies of jumping and flickering, we used the nonparametric, two-tailed Mann–Whitney U test (amplitude: P < 0.001; velocity: P < 0.001; latency: P = 0.08; for all tests conducted: NJ = 21, nJ = 33, NF = 12, nF = 43). Additionally, we compared the variance of the latency for both behaviors using the nonparametric, two-tailed Ansari–Bradley test (P < 0.001, Ansari–Bradley test, NJ = 21, nJ = 33, NF = 12, nF = 43). Statistics were performed using python and the scipy library [1.13.1, (46)]. Tracing experiments. Medio-lateral distributions (“x-percentage”) of ABD, ADD, and small ABD neurons were tested for normality using the Shapiro–Wilk test: ABD MNs were normally distributed (P = 0.20), as were small ABD MNs (P = 0.36), while ADD MNs were not (P < 0.05). Therefore, we compared the medio-lateral distributions of ABD MNs/small ABD MNs and ADD MNs using the nonparametric, two-tailed Mann–Whitney U test (ABD—ADD MNs: P < 0.001, small ABD—ADD MNs: P < 0.001, for all tests performed: N = 5, nABD = 146, nsmall°ABD = 120, nADD = 100). Statistics were performed using python and the scipy library [1.13.1, (46)]. Patch clamp electrophysiology. For patch clamp experiments, significance of differences for the analyzed parameters was tested using the nonparametric Mann–Whitney U test. In boxplots, the central marker (line) indicates the median value, the whisker’s length corresponds to 1.5 times the interquartile range and the bottom and top edges of the box show the 25th and 75th percentiles, respectively. Exact numbers of cells analyzed in each experiment are given in the respective sections of the main text. If not stated otherwise, all values are given as mean ± SEM. Linear and exponential curve fits to data were done using MATLAB (2017b, The MathWorks, MA) and the curve fitting toolbox. Supplementary Material Appendix 01 (PDF) Jumping behavior. High-speed video recording showing a lateral view of a hatchet fish becoming airborne. Original footage was recorded at 2000 fps. First passage shows original footage, the second one shows a 100x slowed down video at 20 fps. Jumping behavior. High-speed video recording showing a hatchet fish becoming airborne, with a view from above. Original footage was recorded at 2000 fps. First passage shows original footage, the second one shows a 100x slowed down video at 20 fps. Jumping behavior. High-speed video recording showing a frontal view of a hatchet fish becoming airborne. Original footage was recorded at 2000 fps. First passage shows original footage, the second one shows a 100x slowed down video at 20 fps. Flickering Behavior. High-speed video recording showing a hatchet fish during flickering behavior from a lateral view. Original footage was recorded at 539 fps. The video has been slowed down to 10 fps. μCT scan of a Carnegiella strigata individual. Video shows reconstructions of ABD (blue), ADD (red) muscles and fish body scan (gray). 3D neuronal projection. 3D-footage of a tracing experiment of p2, showing motoneurons in a cleared whole mount. Example of a single pectoral motoneuron reconstructed using Neurolucida (white). Initial orientation at the start of the video: top = rostral, bottom = caudal.