Evolution of novel sensory organs in fish with legs

Abstract

How do animals evolve new traits? Sea robins are fish that possess specialized leg-like appendages used to "walk" along the sea floor. Here, we show that legs are bona fide sense organs that localize buried prey. Legs are covered in sensory papillae that receive dense innervation from touch-sensitive neurons, express non-canonical epithelial taste receptors, and mediate chemical sensitivity that drives predatory digging behavior. A combination of developmental analyses, crosses between species with and without papillae, and interspecies comparisons of sea robins from around the world demonstrate that papillae represent a key evolutionary innovation associated with behavioral niche expansion on the sea floor. These discoveries provide unique insight into how molecular-, cellular-, and tissue-scale adaptations integrate to produce novel organismic traits and behavior.

Keywords: behavioral evolution; chemosensation; comparative biology; evolutionary innovation; leg-like appendages; limb formation; sea robin; sensory papillae; taste receptors; walking fish.

Figure 1.. Sea robins are fish with sensory “legs” used to locate buried prey.

(A) Sea robins (Prionotus carolinus) are fish with leg-like appendages (arrowheads). (B) Sea robins explored tanks with buried mussels by alternating between swimming (fast) and walking on sand (slow). Representative traces from 10 experiments. (C) Sea robins localized and uncovered buried mussels, or capsules containing mussel extract or <3kDa filtered extract, but not capsules containing control sea water (n = 10 trials each). (D) Schematic of electrophysiological recording from leg-specific spinal nerves. (E) Distal legs responded to mechanical stimulation from von Frey filaments of varying stiffness. n = 4 legs with 4–10 stimulations per filament per leg, p < 0.0001 for comparison of curve plateau by sum-of-square F-test, plateau = 0.8342 to 1.163 (distal) vs. 0.1959 to 0.3132 (proximal). Data represented as the mean ± s.e.m. (F) Legs responded to a broad range of chemicals including common tastants, marine osmolytes, and TRP channel agonists. Heatmap of relative responses from >6 legs. Responses to tastants (2 mM L-alanine) exhibited markedly faster kinetics compared to those elicited by TRP channel agonists (1 mM carvacrol). Recording representative of 6 recordings. Abbreviations: N-Me AAs: N-methylated amino acids, L-AAs: L-amino acids, D-AAs: D-amino acids, SCFAs: short chain fatty acids. (G) Sea robins found capsules containing solutions of single leg agonists but not inert chemicals (n = 10 experiments per chemical). (H) Model of chemical diffusion and surface availability in which shallow capsules of high concentrations produce detectable surface chemicals. (I) Sea robins found mussels buried at shallow depths and higher concentrations. n = 10 trials each. See also Figure S1 and Video S1.

Figure 2.. Legs of digging sea robins are specialized sensory organs.

(A) (left) Prionotus evolans. (right) P. carolinus but not P. evolans exhibited digging behavior and found buried prey (n = 10 mussel trials per indicated prey per species, 4 representative traces of mussel experiments for both species). (B) Distal legs of digging P. carolinus were covered in papillae, which were absent in non-digging P. evolans. (left) brightfield, (middle) scanning electron microscopy, (right) immunofluorescence for neural antigen HNK-1 (green) revealed that leg papillae and ridges are densely innervated in P. carolinus and modestly in P. evolans, respectively. Scale bars (1 mm left, 500 μm middle, 25 μm right). (C) Legs of both species responded to strong mechanical stimulation (4g filament), but only P. carolinus responded to light mechanical stimulation (0.07g filament, n = 41 P. carolinus, 22 P. evolans, p <0.0001, t-test with Welch’s Correction) Data points represent mean signal amplitude averaged over poke duration. (D) Sensation of L-amino acids (2 mM) was unique to digging P. carolinus, while both species responded to betaine (2 mM) and showed little response to choline (2 mM). Representative nerve recordings of >5 legs. (E) Digging P. carolinus legs were ~100X more sensitive to L-amino acids than non-digging P. evolans (responses detected at 100 μM in P. carolinus vs 10 mM P. evolans). n = 4 legs. Data in C and E represented as mean ± s.e.m. See also Figure S1 and Figure S2.

Figure 3.. Sensory neurons mediate leg mechanosensation.

(A) Diagram of sea robin leg neural architecture . (B) Responses to 2 mM L-alanine had faster kinetics compared to those elicited by the Piezo mechanoreceptor agonist Yoda1 (5 μM). Traces representative of 6 replicate recordings. (C-D) Cultured sensory neurons from P. carolinus leg ganglia responded to depolarizing K⁺ (70 mM), the TRP channel agonist carvacrol (100 μM), and weakly to the Piezo agonist Yoda1 (5 μM), but not to appetitive (10 mM betaine, L-alanine, L-proline, D₂ABA) or control chemicals (choline 10 mM) (n = 28 cells, RM one-way ANOVA with Geisser-Greenhouse correction and Dunnet’s multiple comparison test to mean of Choline, $p_{\text{Adj}}=0.6730$ betaine, 0.6913 L-ala, 0.7702 D₂ABA, 0.001, Yoda1, <0.0001 carvacrol, <0.0001 K⁺). Data represented as mean ± s.e.m. (E) Sensory neurons were mechanosensitive and could be separated into three functional populations (n = 26, 8 type 1 with fast desensitization, 11 type 2 with intermediate desensitization, and 5 type 3 with slow desensitization). (F) piezo mechanoreceptor mRNA transcripts were enriched in leg ganglia of both species. Scale: z-scaled normalized counts. (G-H) Leg-specific spinal ganglia were enlarged relative to fin ganglia and possessed an expanded population of trpv1-positive sensory neurons (P. carolinus, trpv1 in green, DAPI in blue, scale bar 500 μm). Large populations of ganglia sensory neurons expressed mechanosensitive piezo1 (magenta) and piezo2 (orange) ion channels (DAPI in blue, scale bar = 200 μm). Images representative of 3 animals per species. See also Figure S2

Figure 4.. Taste receptors in papillae mediate leg chemosensation.

(A) Only the papillae-covered distal leg, but not proximal leg surface of digging P. carolinus responded to appetitive chemicals (2 mM betaine or L-alanine, n = 3, p < 0.0180 betaine, p < 0.0001 L-ala, t-test with Welch’s correction). (B) The taste receptor t1r3 (arrowhead) was the most enriched receptor in the chemosensitive distal leg epithelium. (C-D) t1r3 and t1r2 were expressed in surface epithelial cells of papillae in digging P. carolinus, but absent from non-digging P. evolans, visualized by in situ hybridization (scale bars = 100 μm for left, right, 10 μm for middle, images representative of 3 animals per species). (E) T1r2/T1r3 heterodimers (from P. carolinus) responded to the L-amino acids sensed by digging P. carolinus but not non-digging P. evolans (n = 6). RLU = Relative Light Units. Data in A and E represented as mean ± s.e.m. See also Figure S3.

Figure 5.. Papillae are a major gained trait and facilitate predatory digging behavior.

(A) P. carolinus legs formed weeks after hatching (stained with DAPI, scale bars = 1 mm). dpf = days post fertilization. (B) Legs of larval sea robins initially lacked papillae, which later formed in larger juveniles (scale bars = 1 mm). wpf = weeks post fertilization. (C) Onset of digging behavior correlated with papillae formation in developing P. carolinus sea robins (n = 10 trials per age group, key below indicates approximate timeline with shading beginning at first observation of trait, ANOVA with Tukey’s post-hoc test: 5 vs 6 wk. n.s., 5 vs 7 wk. p < 0.0001, 5 vs 10 wk. p < 0.0001). (D) Dorsal view of an F1 hybrid sea robin produced from crossing female P. evolans and male P. carolinus. (E) Hybrid progeny have leg papillae (top, scale bar = 250μm) and responded to appetitive chemicals similar to their digging parent. Representative image of n = 4. (F) Hybrid sea robins exhibited robust digging behavior to find prey, similar to parental P. carolinus that also have leg papillae (n = 10, p < 0.0001 hybrid or P. carolinus vs P. evolans by ANOVA with Tukey’s post-hoc test). (G) Species with leg papillae (red) represent a restricted clade of the sea robin phylogenetic tree . Species in blue lacked papillae, and species in black were not analyzed. (H) Leg morphology of museum specimens and DAPI stained papillae in P. scitulus (scale bar = 200 μm). (I-J) P. scitulus, which has leg papillae, responded to appetitive L-amino acids, while P. tribulus, which lacks papillae, was insensitive (n = 4 recordings per chemical per species). (K) Papillae-expressing and chemosensitive P. scitulus exhibited robust digging behavior to find prey, while P. tribulus did not dig or find prey (n = 10 trials). Data in C, F, J represented as mean ± s.e.m. See also Figure S4.