Putative photosensitivity in internal light organs (organs of Pesta) of deep-sea sergestid shrimps

Many marine species can regulate the intensity of bioluminescence from their ventral photophores in order to counterilluminate, a camouflage technique whereby animals closely match the intensity of the downwelling illumination blocked by their bodies, thereby hiding their silhouettes. Recent studies on autogenic cuticular photophores in deep-sea shrimps indicate that the photophores themselves are light-sensitive. Here, our results further suggest photosensitivity in a second type of autogenic photophore, the internal organs of Pesta, found in deep-sea sergestid shrimps. Experiments were conducted ship-board on live specimens, exposing the animals to bright light, which resulted in ultrastructural changes that matched those seen in crustacean eyes during the photoreceptor membrane turnover, a process that is crucial for the proper functioning of photosensitive components. In addition, RNA-seq studies demonstrated the expression of visual opsins and phototransduction genes in organs of Pesta tissue that are known to play a role in light detection, and electrophysiological measurements indicated that the light organs are responding to light received by the eyes. The long-sought-after mechanism of counterillumination remains unknown, but evidence of photosensitivity in photophores may indicate a dual functionality of light detection and emission.

Methods
Biological replicates of Parasergestes armatus (n=4) were immediately preserved in RNAlater and stored at -80 °C upon live collection. Organs of Pesta were carefully dissected in chilled RNAlater to prevent degradation of nucleic acids and homogenized in TRIzol reagent (ThermoFisher Scientific). Total RNA was extracted following recommendations in DeLeo et al. (2018) and mRNA libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina. Pippin Prep (Sage Science) was used for size selection on barcoded libraries and all samples (n=4) were sequenced on an Illumina HiSeq4000.

FastQC was used to assess the quality of raw sequencing data and adaptor sequences were trimmed with Trimmomatic v0.36 (parameters: adapter.clip 2:30:10:1:true, crop 135, headcrop 15, trim.leading 3, trim.trailing 3, sliding.window 4:20, min.read.length 36). Rcorrector and BBnorm were used to error-correct reads and normalize read coverage. Tissue-specific transcriptomes were assembled de novo using Trinity v2.6.5 (minimum contig length of 200bp, k-mer size of 23) and contamination was removed using Kraken v1.0 with default parameters and NCBI’s (Refseq) bacteria, archaea, and viral databases. BBduk and dedupe (BBTools suite, available at: http://sourceforge.net/projects/bbmap) were used to remove duplicate transcripts and rRNA. All voucher specimens are archived in the Florida International Crustacean Collection (FICC, HBG #8487-8490). 

The organ of Pesta transcriptome was analyzed using a modified Phylogenetically-Informed Annotation (PIA) Tool which can be used to identify putative visual opsins and phototransduction genes using custom databases (including both visual and non-visual opsins as well as other light detection genes, in precomputed phylogenies which enables the discrimination between false positives and/or paralogous genes following a series of BLAST searches).

To further characterize putative opsins recovered by PIA, these sequences were aligned with PROMALS3D using a reference invertebrate opsin dataset (n=283). This dataset included visual opsins across a range of spectral sensitivities as well as non-visual opsins and related G-protein coupled receptors (GPCR) (see DeLeo and Bracken-Grissom 2020 for a list of all taxa included). Model testing and opsin gene tree reconstruction were done with IQ-TREE using an LG general amino acid replacement matrix, under a FreeRate model with 9 rate categories, and empirical base frequencies (LG+R9+F), based on recommendations from ModelFinder. Support was assessed using 1) a Shimodaira–Hasegawa–like approximate likelihood ratio test (SH-aLRT; 10,000 replicates), 2) an approximate Bayes test and 3) an Ultra-fast bootstrap approximation (UFBoot; 10,000 replicates). False positives aligning with non-visual opsins or outgroups were removed.