The ways in which animals sense the world changes throughout development. For example, young of many species have limited visual capabilities, but still make social decisions, likely based on information gathered through other sensory modalities. Poison frog tadpoles display complex social behaviors that have been suggested to rely on vision despite a century of research indicating tadpoles have poorly-developed visual systems relative to adults. Alternatively, other sensory modalities, such as the lateral line system, are functional at hatching in frogs and may guide social decisions while other sensory systems mature. Here, we examined development of the mechanosensory lateral line and visual systems in tadpoles of the mimic poison frog (Ranitomeya imitator) that use vibrational begging displays to stimulate egg feeding from their mothers. We found that tadpoles hatch with a fully developed lateral line system. While begging behavior increases with development, ablating the lateral line system inhibited begging in pre-metamorphic tadpoles, but not in metamorphic tadpoles. We also found that the increase in begging and decrease in reliance on the lateral line co-occurs with increased retinal neural activity and gene expression associated with eye development. Using the neural tracer neurobiotin, we found that axonal innervations from the eye to the brain proliferate during metamorphosis, with little retinotectal connections in recently-hatched tadpoles. We then tested visual function in a phototaxis assay and found tadpoles prefer darker environments. The strength of this preference increased with developmental stage, but eyes were not required for this behavior, possibly indicating a role for the pineal gland. Together, these data suggest that tadpoles rely on different sensory modalities for social interactions across development and that the development of sensory systems in socially complex poison frog tadpoles is similar to that of other frog species.

Ranitomeya imitator tadpole transcriptome: Thirty R. imitator tadpoles in an unrelated study were anesthetized with benzocaine and euthanized by cervical transection. Eyes, brains, and intestines were placed in RNAlater. RNA was extracted from each tissue using Trizol (ThermoFisher, Waltham, MA, USA) according to manufacturer instructions. Poly-adenylated RNA was isolated from each sample using the NEXTflex PolyA Bead kit (Bioo Scientific, Austin, TX, USA) according to manufacturer instructions. Strand-specific libraries for each sample were prepared using the dUTP NEXTflex RNAseq kit (Bioo Scientific). Libraries were pooled in equimolar amounts after library quantification using both quantitative PCR with the KAPA Library Quantification Kit (KAPA Biosystems, Wilmington, MA, USA) and the fluorometric Qubit dsDNA high sensitivity assay kit (Life Technologies, Carlsbad, CA, USA), both according to manufacturer instructions. Libraries were sequenced on an Illumina HiSeq 2500 over a full flow cell to obtain 1,286,107,683 paired end 250 bp reads.  We used Trinity to de novo assemble the R. imitator tadpole transcriptome (parameters: --seqType fq --SS_lib_type RF --normalize_reads --trimmomatic --full_cleanup). We used blastx to compare contigs in the assembly to proteins in the Uniprot Swiss-Prot database (e-value threshold of 1e-5) and retained contigs with a match to a protein in this database. As we were specifically interested in frog transcripts, we removed contigs that had homology to non-vertebrate proteins in the Uniprot Swiss-Prot database, including microbes, nematodes, and arthropods. Our final R. imitator tadpole draft assembly contained 129,863 contigs (N50=499 bp). We then assessed the completeness of this filtered assembly by examining vertebrate ortholog representation using BUSCO v.4.1.4 (Benchmarking Universal Single-Copy Orthologs). BUSCO estimated our assembly’s completeness to be at 87% with 824 complete single-copy BUSCOS, 2092 complete and duplicated BUSCOs, 271 fragmented BUSCOs, and 167 missing BUSCOs out of 3354 total BUSCO groups searched. We annotated the transcriptome using Trinotate. 

PhosphoTRAP: RNAseq is performed on the total (TOT) input RNA sample and the immunoprecipitated (IP) sample that. RNA samples were purified using a SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Takara, Mountain View, CA, USA), followed by library preparation using the Nextera XT DNA Library Prep kit (Illumina, San Diego, CA, USA), both according to manufactures’ protocols. Pooled equimolar library samples were run on an Illumina Hi-Seq 2500. The R script for phosphoTRAP analysis is included. Illumina reads were aligned to a R. imitator transcriptome. Count data were generated using DESeq2 and analyzed in R using paired t-tests on each transcript between the TOT and IP samples. Fold changes were also calculated for each transcript as log2 (IP count / TOT count). Differentially expressed genes were defined as having a p-value under 0.05 and a fold change greater than 1.5 in either direction. We chose to use paired t-tests of transcripts within each group because this better reflects changes in expression associated with begging and reduces the variation present due to intrinsic variables (developmental stage, hunger, etc) that affect total count data. 

Video of light preference behavior trial: Example video is included of a light preference behavior trial. Tadpole were allowed to acclimate in the chamber for ∼10 min. The small petri dish (tadpole arena) was filled with 40 ml of frog water. Tadpoles were transferred to the middle of the arena and behaviors were recorded for 3 minutes. We then flipped the light arena and recorded for an additional 3 minutes. Behavior was recorded from above with a GoPro Hero 6 camera.