Short toxin-like proteins abound in Cnidaria genomes

Abstract

Cnidaria is a rich phylum that includes thousands of marine species. In this study, we focused on Anthozoa and Hydrozoa that are represented by the Nematostella vectensis (Sea anemone) and Hydra magnipapillata genomes. We present a method for ranking the toxin-like candidates from complete proteomes of Cnidaria. Toxin-like functions were revealed using ClanTox, a statistical machine-learning predictor trained on ion channel inhibitors from venomous animals. Fundamental features that were emphasized in training ClanTox include cysteines and their spacing along the sequences. Among the 83,000 proteins derived from Cnidaria representatives, we found 170 candidates that fulfill the properties of toxin-like-proteins, the vast majority of which were previously unrecognized as toxins. An additional 394 short proteins exhibit characteristics of toxin-like proteins at a moderate degree of confidence. Remarkably, only 11% of the predicted toxin-like proteins were previously classified as toxins. Based on our prediction methodology and manual annotation, we inferred functions for over 400 of these proteins. Such functions include protease inhibitors, membrane pore formation, ion channel blockers and metal binding proteins. Many of the proteins belong to small families of paralogs. We conclude that the evolutionary expansion of toxin-like proteins in Cnidaria contributes to their fitness in the complex environment of the aquatic ecosystem.

Figure 1

Phylogenetic tree of the metazoa. The number of protein sequences for each branch is indicated. Data are retrieved from the NCBI taxonomy database.

Figure 2

Complete proteome annotations. The fraction of the proteins annotated as predicted, hypothetical or putative are shown for Apis mellifera, Hydra magnipapillata and Nematostella vectensis.

Figure 3

Scheme of toxin-like proteins (TOLIPs) discovery. The three major steps in TOLIPs discovery and functional inference are shown. The schematic representation of the histogram of the ClanTox prediction for short proteins is shown. The high confidence predictions are indicated as P2 and P3.

Figure 4

Functional inference for the predictions of TOLIPs from Hydra. TOLIPs that are listed are predicted as P2 and P3. Tandem Repeat (TR) proteins are marked by a star. All other functions are marked by colored arrowheads. XP_002156558.1 carries two functional domains.

Figure 5

Cell modulators from Hydra. (A) The sequence XP_002164320.1 is shown. The segments of the sequence that were excluded from the HHPred comparative models are colored gray. HHPred representation of SIBD domain from Hydra and a structural homologue PDB: 3ZXC_A. The domain of 3ZXC includes a Single Insulin-like Growth Factor-Binding Domain Protein (SIBD-1) from the Central American Hunting Spider Cupiennius salei; (B) Set of secreted proteins and their paralogs. The function of these proteins is unknown. However, the spacing and the number of cysteines along the sequences are conserved (marked red).

Figure 6

High confidence predictions from Nematostella. A list of 80 TOLIPs that were predicted by ClanTox as P3 are shown. Major functions are indicated by the colored bar next to the Cysteine pattern scheme. Neurotoxins (NTx, marked yellow) include proteins that were previously annotated as such. Predicted overlooked neurotoxins are marked blue. The other functions are colored as detailed: Gray, Tandem repeats (TR) proteins; Orange, Extracellular regulation and ligand binding; Brown, Protease inhibitors, mainly represented by the Kunitz domain; Green, homologue to specialized domains from Pfam; Light blue, Calciun modulating domains of adhesion and EGF like; M, metaloprotein; Proteins with exceptionally high number of paralogs are assigned by their number; F marks a fragment that apparently belongs to a long protein. These sequences reflect mistakes in the database assignments. The redundant list includes 159 sequences. Most proteins appear in Refseq and GeneBank and thus appear redundant by the NCBI protein database. Only the non-redundant set is shown.