Flatworms have lost the right open reading frame kinase 3 gene during evolution

All multicellular organisms studied to date have three right open reading frame kinase genes (designated riok-1, riok-2 and riok-3). Current evidence indicates that riok-1 and riok-2 have essential roles in ribosome biosynthesis, and that the riok-3 gene assists this process. In the present study, we conducted a detailed bioinformatic analysis of the riok gene family in 25 parasitic flatworms (platyhelminths) for which extensive genomic and transcriptomic data sets are available. We found that none of the flatworms studied have a riok-3 gene, which is unprecedented for multicellular organisms. We propose that, unlike in other eukaryotes, the loss of RIOK-3 from flatworms does not result in an evolutionary disadvantage due to the unique biology and physiology of this phylum. We show that the loss of RIOK-3 coincides with a loss of particular proteins associated with essential cellular pathways linked to cell growth and apoptosis. These findings indicate multiple, key regulatory functions of RIOK-3 in other metazoan species. Taking advantage of a known partial crystal structure of human RIOK-1, molecular modelling revealed variability in nucleotide binding sites between flatworm and human RIOK proteins.

A ll multicellular organisms studied to date have three right open reading frame protein kinase genes (designated riok-1, riok-2 and riok-3), whereas unicellular organisms have only two (riok-1 and riok-2). Functional studies of riok genes in both unicellular and multicellular model organisms have indicated essential roles for riok-1 and riok-2 in ribosome biosynthesis, chromosome stability and/or cell cycle progression [1][2][3][4][5][6][7][8][9][10][11] . The third gene member, riok-3, has not received much attention, being non-essential, but research has shown that RIOK-3 potentially has multiple roles in cells [12][13][14] . Most research has suggested a role for RIOK-3 in associating with ribosomal proteins of the small ribosomal subunit (SSU) and shown that a lack of RIOK-3 results in a build-up of immature SSUs (18S or pre-40S) in the cytoplasm 10,14 . Additionally, RIOK-3 has been demonstrated to inhibit the activation of the nuclear factor kB (NF-kB) transcription factor by interacting with caspase-10, suggesting a role for riok-3 in the regulation of apoptosis and/or cell survival 12 . Structural information for the RIOK family is limited to the partially solved crystal structure for Hsap-RIOK-1 of humans 11 and three crystal structures for species outside of the animal kingdom, namely Aful-RIOK-1 and Aful-RIOK-2 of Archaeoglobus fuldigus (Archaea) and Cthe-RIOK-2 of Chaetomium thermophilum (thermophilic fungus) 8,9 . In recent studies, we have curated the riok gene family for 12 parasitic roundworms (nematodes), predicted functional domains of parasite RIOK proteins using three-dimensional (3D) structural modelling, and prioritized existing kinase inhibitors for repurposing against RIOK-1 of parasitic nematodes 15,16 .
Logically extending these studies 15,16 , we conducted a detailed investigation of the riok gene family in socioeconomically important parasitic flatworms. Some of these flatworms cause neglected diseases of vertebrates. For example, trematodes (flukes) cause diseases such as schistosomiasis, fascioloiasis, clonorchiasis and opisthorchiasis, affecting hundreds of millions of animals and humans 17,18 ; several of these trematodes are class I carcinogens 19,20 and can predispose to viral infections 21,22 . Additionally, cestodes (tapeworms) cause serious diseases, such as echinococcosis, cysticercosis and coenuriasis, in their intermediate hosts (mammals; e.g., livestock animals and humans) [23][24][25] .
Recent advances in high-throughput sequencing have led to a substantial expansion in the availability of genomic and transcriptomic data for parasitic flatworms (platyhelminths). To date, draft genomes of five trematodes (Clonorchis sinensis, Opisthorchis viverrini, Schistosoma haematobium, Schistosoma mansoni and Schistosoma japonicum) and four cestodes (Echinococcus multilocularis, Echinococcus granulosus, Taenia solium and Hymenolepis microstoma) have been published [26][27][28][29] . There has been no detailed exploration of riok genes/RIOK proteins of flatworms. In the present study, using these datasets, we conducted detailed bioinformatic analyses and characterized riok genes and predicted RIOK proteins at the functional and structural levels for selected parasitic flatworms, and investigated their evolutionary relationships within the animal kingdom. We showed that none of the 25 flatworms studied has a riok-3 gene, a finding that is unprecedented for any multicellular organism. Therefore, we investigated whether (i) riok-3 has been lost from flatworms or gained in ecdysozosoans through a gene duplication event, (ii) riok-3 was lost from flatworms entirely as a result of their unique biology and parasitic life style, and (iii) the loss of riok-3 is the result of a loss of specific kinase/ATPaseregulated pathways linked to ribosomal biosynthesis and/or cell growth. In pursuit of specific drug repurposing or design, we took advantage of a known partial crystal structure of human RIOK-1, and employed a modelling approach to generate and compare the nucleotide binding sites of flatworm and human RIOK proteins.

Results
Detailed characterization of riok genes in flatworms. Based on the characterized riok-1, -2 and -3 genes in Caenorhabditis elegans (WormBase gene identification codes 00019698, 00013688 and 00014012), we identified partial rioks in the draft genomes of nine flatworm species, for which published genomic and transcriptomic data were available. Transcripts were inferred based on predicted amino acid sequences, and matching exons in respective genomes were identified in order to curate full-length genes (Table 1). No fulllength riok genes were identified in S. japonicum, as the quality of the genomic data was insufficient to characterize their gene structures. However, the partial Sjap-RIOK-1 and Sjap-RIOK-2 sequences did encode a putative functional kinase domain and were thus retained for subsequent phylogenetic analyses. The riok-3 gene was not found in any of the flatworm genomes or represented in transcriptomes.
Sixteen full-length riok gene structures were defined for the eight flatworms representing trematodes and cestodes ( Figure 1). For the trematodes, riok-1 had an average gene length of 7,422 bp (63,139 bp; standard error of mean, SEM) and 3 exons. The riok-2 genes were not significantly longer (8,325 bp 6 646 bp), but had twice as many exons (6-7 exons). For the cestodes, on average, riok-1 was significantly shorter (2,357 bp 6 345 bp; p , 0.001) than riok-2 (3,590 bp 6 774 bp). The shorter length of riok-1 usually coincided with reduced exon numbers. One exception was the riok-2 gene of Taenia solium, which was shorter (1,681 bp) than Tsol-riok-1 (1,989 bp). The mean coding length of all flatworm riok genes was 1,444 bp (616 bp). We did not find a significant relationship between genome size and gene length or exon number.
Phylogenetic analyses of flatworm RIOKs. Phylogenetic analyses of inferred and curated protein sequence data were used to establish evolutionary relationships for RIOK-1s and RIOK-2s of flatworms ( Figure 2; Supplementary Data 1-3). The RIOKs of Echinococcus granulosus and Echinococcus multilocularis grouped more closely together than to those of Taenia solium, Taenia taeniaeformis and Taenia asiatica, and all of them grouped together with strong support (pp . 0.99) to the exclusion of RIOKs of Hymenolepis microstoma and Mesocestoides corti (see Figure 2). The RIOKs of Opisthorchis viverrini and Clonorchis sinensis grouped more closely together than to those of Fascioloides magna, Fasciola hepatica and Fasciola gigantica, and all of them grouped together with absolute support (pp 5 1.0), to the exclusion of RIOKs of Schistosoma spp. The two distinct groups of RIOKs of trematodes and cestodes grouped to the exclusion of Crassotrea gigas (a molluscan outgroup). Therefore, RIOK-1s and RIOK-2s grouped according to platyhelminth class. There was a positive relationship between exon number and coding length, and evolutionary distance (P , 0.001).

Conservation of RIOK-3 associating proteins in flatworms.
To explain the identified loss of RIOK-3 from flatworms, bioinformatic analyses were performed on the genomes, transcriptomes and inferred proteomes of all available flatworm species, to identify the presence/absence of RIOK-3 associating proteins involved in (a) the biosynthesis of the ribosomal small subunit (SSU), and (b) the tumor necrosis factor receptor-1 (TNF-R1) signalling pathway. Regarding the former, the bioinformatic analysis identified flatworm orthologs for all known metazoan SSU proteins (n 5 30) and associating proteins (LTV-1, ENP-1, TSR-1, DIM-1, PNO-1 and NOB-1) essential for the cytoplasmic assembly of SSU (Supplementary Data 5). For the latter pathway, a loss of caspase-10 and the upstream TNF-R1 signalling proteins -Fas-associated protein with death domain (FADD) and T tumor necrosis factor receptor type 1-associated death domain (RADD) protein became evident ( Figure 4). Additional gene losses were identified in the NF-kB pathway for the caspase-10-associating receptor-interacting protein kinase (RIPK) and NF-kB-inducing kinase (NIK). In flatworms, we identified TNF-R1 that lacked its cytoplasmic death domains, caspase-3 and caspase-8 ( Figure 4).
Structural analysis and functional predictions of flatworm and host RIOK-1s. The inferred amino acid sequences of Taenia solium (cestode) and Schistosoma haematobium (trematode) RIOK-1 were truncated to match the catalytic kinase domain, for which several structures exist in the protein database (PDB) in their free and ligand-bound forms (Homo sapiens RIOK-1, Archaeoglobus fulgidus RIOK-1 and Archaeoglobus fulgidus RIOK-2). For all known RIOKs, the N-lobe, hinge region and C-lobe that form the minimum kinase domain are predicted to be crucial for kinase/ATPase activity 11,[30][31][32] , and the number and location of their secondary   structure elements are conserved 16 . The flexible loop, a characteristic feature of RIOKs 32,33 comprising a sequence of ,20 amino acids, is located between the end of the P-loop and the adjacent a-helix. For flatworm RIOK-1 proteins, the conserved sequence pattern, based on four cestode and four trematode sequences is: Figure 5). Three-dimensional models of the cestode and trematode RIOKs (with exception of Tsol-RIOK-2), and Hsap-RIOK-2, were produced by threading and iterative template fragment assembly, followed by gene ontology analysis. For Tsol-RIOK-2, a comparative modelling approach guided by a structure-based sequence alignment was chosen and three-dimensional model generated using Aful-RIOK-2 as a template.
In order to structurally appraise the nucleotide-binding sites of flatworm RIOKs, Mg-ADP was manually docked into the sites of all RIOK models produced using the experimental structure of Hsap-RIOK-1 in complex with ADP as a template 11 . As expected, there was a high level of conservation when the nucleotide binding sites of flatworm RIOKs were compared with those of the human host (Supplementary Data 6 and Figure 6). In particular, the pivotal residue coordinating the magnesium ion (Asn329 in Hsap-RIOK-1) is strictly conserved among the six proteins studied. The binding poses shown in Figure 6 result from manual docking into modelled structures and are not refined. However, despite these approximations, very similar binding modes of Mg-ADP in the RIOKs of the three species are expected. The relative orientation and conformation of the modelled ADP molecules are immediately obvious when comparing the three panels in Figure 6 for either RIOK-1 or RIOK-2. Among the three species, the RIOK-2 nucleotide binding sites show few conservative amino acid substitutions (Supplementary Data 7 and Figure 6) and can thus be expected to engage in highly similar interactions with bound ligands. In contrast, a comparison of RIOK-1 models shows that only two positions in the nucleotide binding sites relate to amino acid variations: a conservative variation at V194, and a non-conservative variation at L289 (Supplementary Data 7 and Figure 6).

Discussion
Based on an analysis of existing genomes and transcriptomes, we identified riok-1, -2 and -3 in five early metazoans, such as sponges and cnidarians, in five chordates and in ten ecdysozoans (including nematodes, insects and crustaceans). The presence of riok-3 in molluscs and annelids but its absence from flatworms indicates that the loss of riok-3 is unique to flatworms. A detailed analysis of representatives (n 5 25) of the Phylum Platyhelminthes showed that riok-3 has been lost from both parasitic and free-living flatworms. This information shows that the riok-3 gene loss from flatworms is not related to a parasitic life style, but rather to the unique biology and/or physiology of flatworms. The present study suggests a role for RIOK-3 in ribosomal biosynthesis, and a co-localization and pull-down study of Hsap-RIOK-3 has indicated that RIOK-3 is a component of the small ribosomal unit (40S) biosynthesis complex, and is closely linked to ribosomal proteins ENP-1 and LTV-1 14 . Loss-of-gene function studies in Caenorhabditis elegans and Drosophila melanogaster have shown that, unlike riok-1 and riok-2, riok-3 is not essential for survival and development, but silencing of this kinase/ATPase does reduce the maturation rate of the small ribosomal 40S subunit 10 . Based on high sequence similarity in the kinase core-regions between the RIOK-1 and RIOK-3 32 , riok-3 most likely originated from a riok-1 gene-duplication event. We hypothesize that this gene duplication www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 9417 | DOI: 10.1038/srep09417 event has provided metazoan organisms with the capacity to increase ribosome maturation efficiency, allowing for time efficient protein production and cell cycle progression. Therefore, metazoans expressing RIOK-3 are likely able to increase their protein production capability, providing an evolutionary advantage. However, the question arises as to whether the loss of riok-3/RIOK-3 from flatworms has an impact on ribosomal biosynthesis and overall fitness.
Flatworms, such as the free-living planarian Schmidtea mediterranea are amongst the simplest animals with bilateral symmetry and tissues with distinct organs, and are known for their ability to ''degrow'' (reduce cell numbers and reabsorb reproductive organs) when food is scarce and to regenerate entire body parts 34,35 . This remarkable plasticity is driven by a small number of adult stem cells, called neoblasts -the only planarian flatworm cells that can divide to replace cell loss . Recently, such neoblasts have been described in 36 non-planarian, parasitic flatworms, such as Echinococcus multilocularis 37,38 . Some evidence also suggests that somatic cells of some flatworms do not undergo cell division 34,38 . This apparent lack of cell division in somatic cells is reflected in the reported loss of NIMArelated kinases (NEKs), a family of protein kinases known to play a key role in cell cycle progression 29 , and would result in a reduced protein production pressure on flatworm somatic cells. Therefore, we propose that the loss of the riok-3 gene from flatworms and the likely reduction in ribosomal maturation efficiency have a limited impact on the non-dividing somatic cells of flatworms, not reducing flatworm fitness. To further support this proposal, we studied the presence/absence of RIOK-3-associating proteins in the ribosomal biosynthesis pathway. One would expect the loss of riok-3 in flatworms to coincide with a loss of other SSU proteins. We identified all 30 ribosomal and seven associating SSU proteins in the late ribosomal biosynthesis pathway of the flatworms investigated here. The conservation of the SSU associated proteins among these flatworms, despite the loss of RIOK-3, raises questions as to the primary functional role of RIOK-3. This kinase has also been proposed to be involved in the regulation of NF-kB signalling, which is part of the TNF-R1 pathway 12 . In vertebrates, caspase-10 is one of the initiator enzymes downstream of TNF-R1, and is crucial for the activation of both apoptosis and the NF-kB signalling pathway 39 . Caspase-10 activates the apoptosis pathway through its proteolytic domain, while it activates the NF-kB pathway via its N-terminal death effector domains (DED) 40 . In the NF-kB pathway, the caspase-10 DED domains interact with specific kinases (RIP and NIK), independently of the proteolytic activity of caspase-10, and lead to cell growth and cell differentiation, an opposite response to the apoptosis pathway 40 . The molecular basis for cross-talk between these two pathways is still not well understood. In vertebrates, RIOK-3 has been shown to inhibit the NF-kB pathway, by competing with the kinases RIPK and NIK for the caspase-10 DED domains, independent of its ATPase activity 12 . Lee et al. 41 suggested that the TNF-R1 pathway in flatworms has been lost and that apoptosis is limited to the intrinsic or stress pathway. A bioinformatic search showed the absence of caspase-10 orthologs from the flatworm genomes and transcriptomes studied here, suggesting that the loss of RIOK-3 is linked to those of caspase-10 and/or its target DED domains. In addition to the losses of RIOK-3 and caspase-10, we did not find most other upstream and downstream components of either the apoptosis or NF-kB pathway (Figure 4). However, we did identify TNF-R1 orthologs in flatworms, but these receptors are known to lack dead domains essential for intracellular signalling, supporting the loss of functionality of the extrinsic apoptosis and caspase-10induced NF-kB pathways in flatworms 42 . Based on this analysis, we suggest that the loss of caspase-10 and its associating proteins represent a driving force for the loss of the riok-3 gene from flatworms. In addition, the conservation of all SSU components among flatworms suggests that the regulation of cell death and cell growth through the TNF-R1 pathway represents the primary functional role for RIOK-3 rather than of ribosomal biosynthesis.
The riok genes are receiving increased attention because of their potential as drug targets for anti-cancer and anti-infective compounds 10,15,16,43 . This is particularly true for riok-1 and riok-2, for which loss-of-gene-function studies have shown their essentiality and a likely involvement in vital biological processes, such as ribosomal biosynthesis, chromosome stability and/or cell cycle progression [1][2][3][4][5][6][7][8][9][10][11] . We showed that the riok-1 and riok-2 genes are present and transcribed in 52 metazoans studied here, including 25 flatworms representing trematodes and cestodes. The variability in the observed riok gene structures of four trematodes and four cestodes (exon numbers ranging from 3 to 7, and gene lengths ranging from 1,618 to 14,971 bp) is significantly lower compared with that previously reported for nine nematode species 16 . This variability in gene structure is not related to genome size of species. However, we did observe a positive relationship between the riok gene compactness and the respective evolutionary position of a flatworm species, suggesting relative conservation in riok structure within the trematode and cestode groups.
Various classes of compounds, such as pyrazinoisoquinolines (praziquantel), benzimidazoles (triclabendazole), artemisinins (artesunate), synthetic trioxolanes (closantel) and synthetic peroxides [44][45][46][47][48] , are used to treat parasitic flatworm infections in humans or animals, but some are ineffective against particular developmental stages of some worms 45,47 , and others are limited by a narrow spectrum of activity and/or emergent drug resistance 45,47 . Therefore, there is a need to work toward new and improved drugs against flatworms based on an understanding of their molecular biology. In this context, it is critical to target gene products that are transcribed or expressed in suitably druggable developmental stages of these worms, as is true of RIOKs. The parasitic trematode Schistosoma haematobium displays consistent and moderate transcription levels of the riok-1 and riok-2 genes among developmental stages and between sexes 27 . A similar transcriptional profile is reported for three species of parasitic nematodes (Haemonchus contortus, Brugia malayi and Ascaris suum) 16 and in the vinegar fly, Drosophila melanogaster 49 . The constitutive and moderate riok transcription levels observed across parasitic trematodes, nematodes and insects suggest that RIOKs perform house-keeping functions. This finding agrees with their predicted roles in essential pathways, such as ribosome biosynthesis and cell cycle progression [1][2][3][4][5][6][7][8][9][10][11] .
In the absence of crystal structures of RIOK proteins of flatworms, a molecular modelling approach was employed to compare nucleotide binding sites between parasite and mammalian RIOKs. Since these sites are considered druggable, they are targeted by the majority of current kinase inhibitors for therapeutic uses 50 . The topologies of these binding sites are very similar when compared among different kinases, and the problem of selectivity arises when designing inhibitors specifically against pathogen RIOKs. Selectivity not only relates to a distinction between parasite and host RIOK sites, but also nucleotide binding sites in any of the .500 human kinases. As a first step in this direction, the present study aimed to appraise any suitable variation in the nucleotide binding sites in RIOKs between flatworms (cestodes and trematodes) and a mammal (human). A promising lead may be the residue in position of L289 of human RIOK-1, which is P197 in Schistosoma haematobium and K187 in Taenia solium. The introduction of a basic residue in RIOK-1 of all cestodes and the space alteration of the nucleotide-binding site in orthologs in the case of trematodes are points of interest for the design of ligands that specifically target RIOK-1 proteins of flatworms, as opposed to those of the host.

Methods
Gene prediction and identification. From public databases, we extracted genomic, transcriptomic and protein datasets for the identification, isolation and curation of fulllength riok genes from the trematodes Clonorchis sinensis (GenBank Assembly ID: GCA_000236345.1), Opisthorchis viverrini (GenBank Assembly ID: GCA_000236345.1), Schistosoma haematobium (NCBI BioProject ID: PRJNA78265), Schistosoma mansoni (GenBank Assembly ID: GCA_000237925.2) and Schistosoma japonicum (GenBank Assembly ID: GCA_000151775.1), and the cestodes Echinococcus multilocularis (GenBank Assembly ID: GCA_000469785.1), Echinococcus granulosus (GenBank Assembly ID: GCA_000524195.1), Taenia solium (NCBI BioProject ID: PRJNA183343) and Hymenolepis microstoma (GenBank Assembly ID: GCA_000469805.1). Using transcripts inferred from partial amino acid sequences of RIOK proteins, riok genes were located in individual genomes and extracted using the  ). Posterior probabilities (pp) were calculated using 2,000,000 generations and the last 25% of these generations were used for phylogenetic analysis, employing four simultaneous tree-building chains, saving every 100 th tree. For each of the RIOK subfamilies found, a phylogenetic analysis of amino acid sequence data for the minimum kinase domain was conducted. RIOK sequences from Crassotrea gigas (a mollusc; GenBank Assembly ID: GCA_000297895) were used as outgroups; evolutionary distances were calculated using the toolkit BioPhylo 53 . Structural characteristics of individual genes were established (e.g., intron length, coding length, gene length, number of exons and genome size). General linear modelling in R was used to detect any relationships between evolutionary distance and gene structural characteristics for each RIOK subfamily.
Structural analysis of flatworm and host RIOK-1 proteins. Structure-based amino acid sequence alignments were generated using the program SBAL 54 , and conserved and novel domains identified. The structure of Hsap-RIOK-1 was studied using the crystal structure deposited in the PDB (accession code 4otp) 11 . Three-dimensional models were built for truncated RIOK proteins from representative flatworms (RIOK-1 of Taenia solium, and RIOK-1 and RIOK-2 of Schistosoma haematobium) and their host species (RIOK-2 of H. sapiens), using the threading and iterative template fragment assembly approach, implemented in I-TASSER 55 . Models with a positive C-score (i.e. confidence score for estimating the quality of the predicted model; ranging between -5 and 2) and the highest TM-score (TeMplate quality-score of the generated 3D model and a known RIOK crystal structure) were selected. For Tsol-RIOK-2 (Taenia solium), a homology model was generated using the program MODELLER 56 employing the crystal structure of Aful-RIOK-2 (Archaeoglobus fulgidus; PDB accession code 1zar) as a template. Briefly, a secondary structure-based amino acid sequence alignment of Tsol-RIOK-2 and Aful-RIOK-2 was prepared using SBAL 54 and used to guide comparative modelling calculations. Twenty independent models were calculated, and that with the lowest energy was selected. All models were superimposed on the structure of Hsap-RIOK-1, and ADP docked into the five modelled structures by manual rigid-body docking using the program Coot 57 . An appraisal of the nucleotide binding site and possible protein-ligand interactions was undertaken by visual inspection of structures in Coot as well as distance analyses using in-house scripts (AH) from PCSB software 58 . Sequences of domains that displayed differences in the alignment between parasitic flatworms and human RIOK-1s were compared using MUSCLE-guided alignments 59 and displayed using the program WebLogo 60 .