Functional kleptoplasty is a photosymbiotic relationship, in which photosynthetically active chloroplasts serve as an intracellular symbiont for a heterotrophic host. Among Metazoa, functional kleptoplasty is only found in marine sea slugs belonging to the Sacoglossa and recently described in rhabdocoel worms. Although functional kleptoplasty has been intensively studied in Sacoglossa, the fundamentals of the specific recognition of the chloroplasts and their subsequent incorporation are unknown. The key to ensure the initiation of any symbiosis is the ability to specifically recognize the symbiont and to differentiate a symbiont from a pathogen. For instance, in photosymbiotic cnidarians, several studies have shown that the host innate immune system, in particular scavenger receptors (SRs) and thrombospondin type 1 repeat protein superfamily (TSRs), is playing a major role in the process of recognizing and differentiating symbionts from pathogens. In the present study, SRs and TSRs of three Sacoglossa sea slugs, Elysia cornigera, Elysia timida, and Elysia chlorotica, were identified by translating available transcriptomes into potential proteins and searching for receptor specific protein and/or transmembrane domains. Both receptors classes are highly diverse in the slugs, and many new domain arrangements for each receptor class were found. The analyzes of the gene expression of these three species provided a set of species-specific candidate genes, i.e. SR-Bs, SR-Es, C-type lectins, and TSRs, that are potentially relevant for the recognition of kleptoplasts. The results set the base for future experimental studies to understand if and how these candidate receptors are indeed involved in chloroplast recognition.

Publicly available assembled transcriptomes of the Sacoglossa Elysia cornigera: NCBI TSA version GBRW00000000.1; E. timida: TSA version GBRM00000000.1; and E. chlorotica: http://cyanophora.rutgers.edu/Elysia-expression/) were first clustered using CD-HIT v4.6.8 with default parameters (Fu, Niu, Zhu, Wu, & Li, 2012; Li & Godzik, 2006) and then translated into the longest open reading frame to retrieve potential proteins using TransDecoder v3.0.1 (Haas & Papanicolaou, 2015) with default settings. The datasets were subsequently subjected to a BLASTP search against the UniProt database version 11/13/19 (The UniProt Consortium, 2019) setting the E-value to 1e^-10. Taxonomic assignment for each protein sequence was performed using the UniProt taxonomic database and sequences were subsequently filtered for Metazoa annotations (Supporting Information 1). Respective short reads were downloaded for each species from the short read archive deposited in GenBank (Elysia chlorotica; NCBI SRA sample accession SRS3101883), Elysia timida SRS706683, and Elysia cornigera SRS706681). Reads were then mapped using Bowtie2 v2.3.4.3 (Langmead & Salzberg, 2012) onto the clustered transcriptomes. Transcript abundance of sequences with a raw read count of at least 100 raw counts in any two samples tested was estimated using RSEM (Li & Dewey, 2011) implemented in Trinity v.2.9.0 (Grabherr et al., 2011). Differential gene expression analyzes were performed using edgeR v3.30.3 (Robinson, McCarthy, & Smyth, 2010). For feeding juveniles of E. chlorotica, specimens fed for five days were compared to the apo-symbiotic state (initial); specimens fed for seven days with specimens fed for five days (transient); and specimens fed for 10 days with specimens fed for seven days (stable). For E. timida and E. cornigera the freshly fed animals were compared to the different starvation periods. Only genes with a log₂ fold change (L2FC) > 1 or < -1 were considered as significantly differentially expressed, because for E. cornigera and E. timida no biological replicates were available. The domain architecture of the filtered protein sequences were characterized by using HMMER v.3.1b2 (Eddy, Wheeler, & the HMMER development team, 2015) with default settings against the protein database PfamA 31.0 (Finn et al., 2016). Transmembrane regions (TM) were identified using the TMHMM server v.2.0 (Sonnhammer, Von Heijne, & Krogh, 1998; Krogh, Larsson, Von Heijne, & Sonnhammer, 2001). Sequences were then filtered for the different receptor class specific domains, as defined in PrabhuDas et al. (2014, 2017).