The parasite Schistocephalus solidus secretes proteins with putative host manipulation functions

Background Manipulative parasites are thought to liberate molecules in their external environment, acting as manipulation factors with biological functions implicated in their host’s physiological and behavioural alterations. These manipulation factors are part of a complex mixture called the secretome. While the secretomes of various parasites have been described, there is very little data for a putative manipulative parasite. It is necessary to study the molecular interaction between a manipulative parasite and its host to better understand how such alterations evolve. Methods Here, we used proteomics to characterize the secretome of a model cestode with a complex life cycle based on trophic transmission. We studied Schistocephalus solidus during the life stage in which behavioural changes take place in its obligatory intermediate fish host, the threespine stickleback (Gasterosteus aculeatus). We produced a novel genome sequence and assembly of S. solidus to improve protein coding gene prediction and annotation for this parasite. We then described the whole worm’s proteome and its secretome during fish host infection using LC–MS/MS. Results A total of 2290 proteins were detected in the proteome of S. solidus, and 30 additional proteins were detected specifically in the secretome. We found that the secretome contains proteases, proteins with neural and immune functions, as well as proteins involved in cell communication. We detected receptor-type tyrosine-protein phosphatases, which were reported in other parasitic systems to be manipulation factors. We also detected 12 S. solidus-specific proteins in the secretome that may play important roles in host–parasite interactions. Conclusions Our results suggest that S. solidus liberates molecules with putative host manipulation functions in the host and that many of them are species-specific. Graphical abstract Supplementary Information The online version contains supplementary material available at 10.1186/s13071-021-04933-w.

of parasites infecting humans [12] and livestock [13] have been described, both in terms of protein composition and function [14,15] (see Table 1 for a review).
The secretomes that have been examined so far are enriched in peptidases and proteases [12,15], which are known to weaken the host immunity barriers. Other secreted proteins, such as paramyosin in the blood fluke Schistosoma mansoni, have been shown to help the parasite to escape the host immune response, while secreted proteins involved in calcium functions have impacts on the host neural activity [12]. In the context of behavioural manipulation, the secretome is a logical potential source of manipulation factors. However, the secretome content of a behaviour-manipulating parasite has rarely been investigated, to the point that secretomes are referred to as "the missing link in parasite manipulation" [7]. The literature contains several reports from which it is possible to infer a list of putative manipulation factors, which would target the neural and the immune systems of the hosts and induce behavioural changes (Table 1). Our knowledge regarding if and how many proteins with neural and immune functions can be found in the secretomes of manipulative parasites is very limited, and is based in many cases on inferred proteins rather than actual detection.
One particularly powerful model to study behavioural manipulation is the cestode Schistocephalus solidus [27]. This parasite exhibits a complex life cycle based on trophic transmission that includes three hosts: a copepod, the threespine stickleback (Gasterosteus aculeatus, obligatory intermediate fish host), and a fish-eating bird, in which S. solidus reproduces [27,28]. S. solidus infects the threespine stickleback's abdominal cavity through the ingestion of a parasitized copepod [29]. The Table 1 Proteomic content (directly measured or inferred) of the secretomes of different parasite species (1) Species of human importance. (2) Species studied in the context of behavioural manipulation. a Proteins inferred from experimental or bioinformatics evidence, i.e. consequences of the infection by S. solidus on the threespine stickleback's morphology [30], physiology [31], immune system [32], and behaviour [33] are well documented. For example, sticklebacks infected by S. solidus show drastic behavioural changes that result in a loss of the anti-predator response [27]: infected fish are more exploratory [34], less anxious [35], and bolder in the presence of a predator [36] than non-infected fish.
Most of these behavioural alterations seen in infected fish appear after several weeks, when the worm has grown to reach the infective stage within its intermediate host (i.e. larger than 50 mg) [37]. The infective stage coincides with the time at which S. solidus is ready to reproduce in the bird [27,38], which also generally corresponds to the activation of the immune response in the host. In the first phase of infection, the adaptive immune response is generally not activated in the fish. It is only when the worm reaches the infective stage that an ineffective upregulation of the respiratory burst activity occurs [32]. Nevertheless, activation of the immune response through the production of reactive oxygen species (ROS) by granulocytes has been shown to occur early during infection, depending on the genotype of the stickleback population [39]. Several studies have suggested that the manipulation of the stickleback's behaviour increases the worm's transmission rate to its final avian host [27]. Yet, the adaptive value for S. solidus of such behavioural changes has never been demonstrated [40], and it is possible that these behavioural modifications in the fish may solely result from a side effect of infection [7], such as the effect of the parasite mass burden or of the activation of the host immune response [35,41]. To demonstrate that behavioural changes in the host are the result of direct parasitic manipulation, the first step is to determine whether the parasite can liberate molecules in its external environment, and if yes, to study their functions in relation with the host's phenotype perturbations.
This host-parasite system was used for the first experimental demonstration of a complex life cycle of a trophically transmitted parasite in 1790 (as reviewed in [27]). A rich body of work on this model system during the past 50 years has shown, using enzymatic assays, that the activity of proteases [42] and transferases [43] is required for S. solidus survival and growth. Furthermore, a partial genome of the worm is available [44], and extensive transcriptome data has been produced [45,46]. Quantification of the transcriptome dynamics across the life stages has uncovered that when the worm reaches the infective stage in its fish host, genes involved in neural pathways and sensory perception have higher expression levels compared to the earlier stages in the same host, which are characterized by upregulation of growthrelated pathways [46]. Furthermore, vesicles are present inside the S. solidus tegument, as shown through scanning and transmission electron microscopy [47]. Moreover, S. solidus excretes (through passive mechanisms) or secretes (through active mechanisms) molecules, including (uncharacterized) proteins, and these secretions are sufficient to affect its fish host behaviour [48]. Finally, a well-annotated genome of the threespine stickleback host is also available [49], which is important to adequately differentiate proteins coming from the host and the parasite. The Schistocephalus-stickleback system is thus appropriate to test the presence of manipulation factors. However, the nature of the protein content of the S. solidus secretome has never been explored [7,48]. Analysing the proteomic fraction of the secretome of S. solidus, and its potential enrichment in manipulation factors involved in neuronal and immune functions, will help us to understand if the behavioural changes of the host could be induced by parasitic manipulation through the secretome.
Here, we characterized the worm's whole-body proteome and the protein fraction of the secretome of S. solidus using liquid chromatography-tandem mass spectrometry (LC-MS/MS) [50]. We focused on individuals in the infective stage of development so that the secretome may include manipulation factors that could be associated with the fish host's behavioural changes. In helminth parasites, a portion of proteins from the proteome are passively released in the external environment as metabolic waste products, contributing to an important fraction of their secretome content [51]. Therefore, the secretome is generally a subset of the proteome in terms of protein content [52]. We thus expected that the secretome of S. solidus would also mainly be a subset of its proteome. However, in the context of behavioural manipulation, we hypothesized that proteins could also be actively liberated by the parasite in its external environment [51], so that they would be enriched in the secretome compared to the proteome. Based on what has been described in previous parasitic systems (references reviewed in Table 1), if S. solidus manipulates stickleback behaviour with its secretions as we hypothesize, then its secretome would include proteases as well as proteins with neural and immune functions. Because the worm is not in direct contact with the vertebrate host brain [27], we also expected to detect proteins involved in cell communication, cell-cell signalling, or transport functions. These proteins would mediate the communication of the worm with its host's brain to induce potential neural and immune changes, and ultimately behavioural alterations.
As the genome of S. solidus is used during LC-MS/ MS as a reference database, a more thorough annotation of the genome than the one publicly available [44] could allow us to detect more proteins, including potential manipulation factors. We therefore used a multipronged approach combining genomics with proteomics (described in Fig. 1). We first sequenced the genome of S. solidus using a combination of long and short reads to combine longer contigs and improved annotation. Then, we investigated the global proteomic composition of the proteome and secretome of S. solidus using LC-MS/MS.

Genome sequencing of Schistocephalus solidus Worm collection
We caught threespine sticklebacks from Lac Témiscouata (Québec, 47°40′33″N 68°50′15″W), where fish are known to be infected by S. solidus [35], using minnow traps in July 2015. Fish were brought to the Laboratoire Aquatique de Recherche en Sciences Environnementales et Médicales at Université Laval (Québec, Canada) and were maintained under a 12 h:12 h light/dark cycle and a water temperature of 15 °C. Fish were fed daily with a mixture of blood worms and Artemia. After 10 months, one fish (out of 40) exhibited the morphological changes typically induced by S. solidus [30] and was consequently sacrificed to collect the worm. The fish was euthanized with an overdose of MS-222 (75 mg/l mg/kg) and dissected to confirm the infection by S. solidus. The worm was immediately put in ethanol 90% and stored at 4 °C until genome sequencing. All the other fish were later used during a behavioural experiment [35].

DNA extraction and sequencing
Genomic DNA was extracted using the DNeasy Blood and Tissue kit (Qiagen Inc., Valencia, CA, USA) from ~ 20 mg of tissues. After lysis, we added 4 µl of RNase A (10 mg/ml). Elution was done twice in 100 µl of elution buffer. To reach the desired concentration for the Nanopore library preparation and Illumina sequencing, we concentrated DNA with a SpeedVac concentrator (Thermo Fisher, Waltham, MA, USA) for 1 h.
Illumina HiSeq X sequencing The DNA libraries for Illumina sequencing were prepared using a shotgun PCRfree library preparation (Lucigen) Illumina Library at the McGill University and Genome Quebec Innovation Centre (Montréal, Canada). Sequencing was performed at the same centre on an Illumina HiSeq X instrument, using paired-end reads (PE150). A total of 157,475,128 reads were obtained. The estimated genome size is 540 MB [44].
Oxford Nanopore Technologies (MinION) sequencing Library preparation for Oxford Nanopore sequencing was done with a PCR-free ligation sequencing kit SQK-LSK108 (ONT, Oxford, UK). Briefly, approximately 2.5 µg of high-molecular-weight DNA was repaired using

Genome assembly and annotation
All fastq ONT files were pooled in one file prior to assembly. ONT sequences were assembled using Flye v2.4 [53,54]. The final assembly produced 17,882 scaffolds, the largest scaffold being 919,337 nt, with an N50 of 121,189 nt. The mean coverage across scaffolds was 20×. A first phase of correction (polishing) was carried out with nanopolish (v0.10.2) (https:// github. com/ jts/ nanop olish). A total of 3,473,881 changes were applied. A second correction phase with one Illumina HiSeq X paired-end sequence library (20× coverage) was also done with Pilon (v1.22) [55]. A total of 4,889,938 changes helped to improve the quality of the assembly sequences. Scaffolds that had an average coverage of less than 10× coverage and those that were shorter than 500 bp were removed because they were of limited use prior to genome annotation. These represented less than 1% of the data, and all contigs that contained hits with the transcripts from S. solidus [46] remained after this selection. This left a total of 15,357 scaffolds and 625,207,408 nucleotides. This Whole Genome Shotgun project has been deposited at DDBJ/ENA/GenBank under the accession WEDG00000000.1 and the assembly ID ASM1759139v1.
Completeness of the genome assembly We used a dataset of 24,765 transcripts from S. solidus that we previously published [46] and mapped them on the de novo assembly using GMAP (v2019-03-15) [56] as implemented in A multi-pronged approach to study the secretome of Schistocephalus solidus In Fig. 1 The multipronged approach designed to detect potential manipulation factors in the secretome of Schistocephalus solidus. On the left, the proteomic approach based on LC-MS/MS aims at describing the global proteomic composition of the proteome and the secretome of S. solidus. On the right, the genomic approach aims at producing a novel genome sequence and assembly of S. solidus to improve protein coding gene prediction and annotation for this parasite. The genome, whose quality is improved, can be used as a new reference database to infer proteins in the proteome and the secretome of S. solidus. Combined, these approaches allow us to characterize the secretome of S. solidus at the infective stage, including the uncovering of proteins detected only in that fraction and not in its proteome, thus representing potential manipulation factors the pipeline GAWN v0.3.2 (https:// github. com/ enorm andeau/ gawn). An assembly quality analysis was performed using the Benchmarking Universal Single-Copy Orthologs (BUSCO), and BUSCO groups were searched in the Metazoa database [57].
Protein coding gene prediction To find out the proportion of repeated regions in the genome and to obtain a masked assembly prior to running BRAKER2 to predict protein coding genes, we built a RepeatModeler (1.0.8) database [58] based on the new genome sequence and ran RepeatMasker (4.0.6) based on that database [59]. We used BRAKER2 [60][61][62][63][64][65] for protein coding gene prediction using two approaches. The first approach was ab initio as it was not based on external data to find open reading frames (ORFs) but only on genes predicted by GeneMark-ES that are selected for training Augustus [60,61]. The second approach used the alignment (bam files) of the transcripts from S. solidus [46] on the genome. The two sets of ORFs obtained with the two approaches were merged, and duplicate sequences were removed, which allowed us to obtain a final number of predicted ORFs. We used a BLAST analysis to quantify the percentage of the S. solidus transcriptome we previously published [46] that was found in the genes predicted using Augustus and vice versa.
Identification of sequences specific to the new assembly The predicted ORFs were locally aligned using BLAST+ [66,67] against a database of 43,058 protein sequences from S. solidus obtained from release 230 of the NCBI (March 2019, https:// www. ncbi. nlm. nih. gov). From the predicted ORFs, we selected those that had no BLAST match or that had no significant match based on the fact that the length of the alignment was less that 80% of the length of the query or that had less than 90% of identical nucleotides over the length of the query [68]. These unmatched sequences that were specific to the assembly were used as one of the reference databases during LC-MS/MS analysis (see "Protein identification").

Mass spectrometry characterization of the worm proteome and secretome Worm and secretome collection for mass spectrometry analysis
Sampling of experimental individuals Whole worms were collected from wild-caught fish acclimatized to laboratory conditions. Juvenile sticklebacks came from Lac Témiscouata (Québec, 47°40′33″N 68°50′15″W), the same lake that was used to collect a worm for genome sequencing. Fish were caught using a seine in August 2016. They were brought to the Laboratoire Aquatique de Recherche en Sciences Environnementales et Médicales at Université Laval where they were raised for 1 year in 80-l tanks under a light/dark photoperiod of 12 h:12 h and a temperature of 15 °C reflecting their natural environment conditions (Québec, Canada). Fish were fed daily with brine shrimps.

Collection of proteome and secretome samples
In summer of 2017, 51 fish were individually isolated in 2-l tanks. The day following isolation, fish were injected with 100 µl of phosphate-buffered saline (PBS; pH 7.4, Life Technologies) in their abdominal cavity in order to sample their fluids to detect infection by S. solidus following the method described in [72]. This protocol was repeated the next day. If fish were detected as infected, they were euthanized with an overdose of MS-222 (75 mg/l mg/kg) and dissected to confirm the infection by S. solidus and to collect the worm and its secretome. Fish sex, size, and mass, and S. solidus mass and number in each fish were noted. We found that five fish were infected, each harbouring a worm whose weight was above 50 mg (worm 1 = 485.3 mg; worm 2 = 504.1 mg; worm 3 = 286.5 mg; worm 4 = 544.5 mg; worm 5 = 220.9 mg). All the worms used to collect proteome and secretome samples were far above the mass threshold of 50 mg, which is commonly used to define the parasitic infective stage that coincides with the appearance of the behavioural changes in the fish host [27,37,38]. Furthermore, behavioural changes in fish infected by worms of similar masses were previously reported in the Témiscouata population [48]. Therefore, we considered these worms and secretome samples to be representative of the manipulative stage of S. solidus. We only selected fish that were infected by a single worm to prevent potential effects of multiple infections on the proteomic content of the secretome. The worm secretome was collected according to a protocol adapted from [73]. Each worm was rinsed with 1 ml of PBS (pH 7.4, Life Technologies) to remove fish fluids and then immediately put in a 2-ml tube of PBS. The tube was covered with aluminium foil to protect the worm from light and placed in a water recipient at the same temperature as the fish tanks (15 °C) for 2 h. The worm was removed from the tube, and a tablet of complete Mini Protease Inhibitor (Sigma) was added to the tube to protect the proteins from the protease activity. The worm tissues were then snap-frozen in liquid nitrogen and stored dry at −80 °C for proteomics analysis. The liquid in which the worm was incubated (PBS + potential secretome collected) was stored at −20 °C. Further experiments were therefore conducted with five worms and their respective secretome.

Preparation of worm tissues and secretome for in-gel digestion
Worms were individually washed with 2 ml of PBS (pH 7.4, Life Technologies) to remove potential remaining contaminants from the fish. They were then cut into three equal pieces that were each put into a tube containing 700 µl of lysis buffer (4% (w/v) sodium dodecyl sulfate (SDS) in 100 mM Tris, pH 8-10, mM dithiothreitol [DTT]). Six sterile ceramic beads (size 2.8 mm) were added to each lysis tube, and the samples were homogenized for 20 s at 6000 rpm using a Precellys homogenizer (Bertin Technologies). Homogenization was repeated three times. Samples were put on ice for 1 min between each run. Samples were then spun at 10,000×g at 4 °C for 10 min. For each individual worm, the three homogenates obtained were pooled together and redistributed into equal volumes into two tubes. Samples were heated at 95 °C for 10 min, then spun for 10 min at room temperature. The supernatant of each sample was collected into a new tube, and protein concentration was calculated using the extinction coefficient measured with a NanoDrop 1000 spectrophotometer (A280 nm; Thermo Scientific). The concentration obtained for each worm lysate was respectively: 24.14 mg/ml; 22.03 mg/ml; 15.66 mg/ml; 28.53 mg/ml; and 11.37 mg/ml. Worm lysates were kept at −20 °C before performing in-gel digestion.
For the secretome, the protein concentration of each of the liquids in which the worms were incubated ("secretome" samples) was measured using a Pierce Coomassie (Bradford) Protein Assay Kit (Fisher Scientific). Bovine serum albumin (BSA) at 2 mg/ml was used as standard (seven-point standard curve ranging from 0 to 25 µg/ml). The concentrations obtained were: 13.11 µg/ml; 15.09 µg/ml; 9.83 µg/ml; 16.59 µg/ml; and 7.46 µg/ml. For each secretome sample, 10 µg of proteins were precipitated with trichloroacetic acid (TCA) [74]. Precipitated samples were directly used for in gel-digestion.

In-gel digestion
For each sample, 50 µg of proteins from worm lysate and 10 µg of proteins from secretome were resolved on a 12% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) with a BenchMark protein ladder (Invitrogen) and a negative control (SDS-PAGE loading buffer and water). Migration was performed during 60 min at 175 V for the worm lysates, and during 30 min at 175 V for secretomes. Coomassie blue G250 (PMID: 15,174,055) was used for staining overnight (Additional file 1). Each migration lane was cut into five fractions for worm lysates and three fractions for secretomes. In-gel digestion was performed on these fractions according to a previously developed protocol [75]. Briefly, after cutting the gel into slices, proteins were reduced using 10 mM dithiothreitol (DTT) for 45 min at 56 °C, then alkylated with 55 mM iodoacetamide (IAA) for 30 min at room temperature in the dark. Digestion was performed overnight at 37 °C with trypsin (Promega V5113; 0.1-1 µg of trypsin depending on the gel staining intensity). The next day, peptides were extracted using an organic solvent (100 µl acetonitrile) and a step-wise protocol (40% acetonitrile then 100% acetonitrile), and dried (< 50 μl). Following in-gel digestion, a STAGE-TIP protocol using C18 extraction disks (3M Empore) was performed to desalt the samples [76,77]. Samples were acidified with trifluoroacetic acid (TFA) (pH < 2.5) before being passed through the tip. At the end of the STAGE-TIP protocol, samples were dried completely using a Vacufuge Plus concentrator (Eppendorf ) and stored at −20 °C until LC-MS/MS analysis.

LC-MS/MS analysis
The peptides were analysed using a quadrupole time-offlight mass spectrometer (Impact II; Bruker Daltonics) coupled to an Easy-nLC 1000 high-performance liquid chromatography (HPLC) system (Thermo Fisher). More information about the LC-MS/MS analysis can be found in the supplementary information (Additional file 2).

Protein identification
We searched the detected mass spectra against the S. solidus genome using MaxQuant (version 1.6.1.0) [78]. Two searches were independently performed: the first search used the larger proteome of S. solidus (43,058 entries, downloaded June 21, 2018, and updated May 6, 2019) from the Universal Protein Resource release 2018-05 and 2019-03 (UniProt https:// www. unipr ot. org/) as a reference database, which includes proteins predicted from the partial genome of S. solidus that was then currently available [44], as well as proteins predicted from the de novo transcriptome [45,46]. For the second search, we used the unmatched sequences that we reported to be specific to our genome assembly (36,140 sequences, see Results) as a reference database. The search included common contaminants and variable modifications of methionine oxidation, and N-acetylation of the proteins. The data was filtered for matches passing a 1% false discovery rate set by MaxQuant [78]. We included the larger proteome of the threespine stickleback host (29,032 entries, downloaded June 21, 2018) from the Universal Protein Resource release 2018-05 (UniProt https:// www. unipr ot. org/) in the search in order to include proteins originating from the fish host during the MaxQuant data search. Subsequent analyses were performed with all the proteins inferred using the larger proteome of S. solidus and the unmatched sequences of the new genome as reference databases.

Data analysis of the proteome
Data analysis was performed with Python custom scripts (version 3.6.4) using Jupyter notebooks (version 5.4.0). A template of the custom script is available in the supplementary information (Additional file 3). Proteins detected in each worm sample were retrieved using MaxQuant. We filtered out of the dataset protein IDs that were solely attributed to the threespine stickleback (Additional file 4), protein IDs with REV coding (reverse hits for false discovery rate filtering), and protein IDs with CON coding (contaminant hits that were added into the search). In some cases, several protein IDs were found by MaxQuant for a specific protein because of high sequence similarities between them. Research on NCBI (https:// www. ncbi. nlm. nih. gov/) and UniProt (https:// www. unipr ot. org/) databases demonstrated that the protein IDs detected for one protein were probably isoforms with identical functions. These multiple protein IDs were kept during annotation to obtain exhaustive functional information, but only the first ID of each protein was kept for enrichment analyses to avoid an overrepresentation of specific processes or functions (see below). This final dataset was used to describe the global composition of the proteome of S. solidus.
We performed two distinct enrichment analyses for the proteins detected in at least one worm sample and for the proteins detected in all worm samples. For proteins with several protein IDs, we kept only the first ID in order to prevent enrichment bias during analysis. Enrichment analysis was performed using the tool FunRich (version 3.1.3) [79]. We constructed a custom reference database using the protein IDs and the Gene Ontology (GO) annotation (biological process, cellular component, molecular function) of all the proteins described in the larger proteome of S. solidus available on UniProt (https:// www. unipr ot. org/ 43058 entries). P-values for enrichment were obtained with a hypergeometric test corrected with the Bonferroni method.

Data analysis of the secretome
Data analysis was performed with Python using the same approach as for the proteome. We separated proteins into two categories: proteins that were shared between the proteome and the secretome samples, and those that were found only in a secretome sample. For the proteins that were shared between the proteome and the secretome samples, we performed two distinct enrichment analyses: one for the proteins detected in at least one secretome sample and one for the proteins detected in all secretome samples. The enrichment analysis was performed using the FunRich tool (version 3.1.3) [79] using the same approach as for the proteome.
We investigated the annotation of proteins found only in the secretome and not the proteome using a three-step approach. During the first step, the protein ID obtained with the MaxQuant analysis was searched in UniProt to retrieve its information if available (protein name and sequence, corresponding coding gene, protein function, localization, and/or structural information). During the second step, we inferred the function of the protein based on sequence homologies using the Basic Local Alignment Search Tool (BLAST) on NCBI (https:// blast. ncbi. nlm. nih. gov/ Blast. cgi). We used BLASTP (protein-protein) using the non-redundant protein sequences (nr) database (August 2018) and TBLASTN (protein-translated nucleotide) using the nucleotide collection (nr/nt) database (August 2018). During the last step, when no information was found with the previous steps or to confirm the information previously found, we used the Pfam database (version 32.0) [70] to infer the function of the protein based on domain organization homologies. In some cases, several protein IDs were found by MaxQuant for a specific protein because of high sequence similarities between these IDs. We found by applying the previous approach that in all of these cases, the putatively redundant protein IDs detected for one protein were indeed isoforms with identical functions. These multiple protein IDs were nevertheless kept during annotation (see below) to obtain exhaustive functional information.

Annotation of proteins
In order to obtain an exhaustive annotation of proteins, we conducted complementary approaches based on sequence and structure analyses (as described below) [56,[80][81][82][83][84][85][86][87][88][89][90] for the proteins detected in the proteome and/or enriched in the secretome for which few or no annotation was available. These analyses were also performed for the proteins that were detected using the unmatched sequences specific to our genome assembly as a reference database, for which limited or no annotation was available.

Protein structure analysis
To construct 3D models of proteins, we searched their homologs in the Protein Data Bank (PDB) database using BLASTP, Delta-BLAST, and HHpred [83]. Then, 3D structure models were built by homology modelling based on their homologous structures (PDB ID: 4xp4_A) using MODELLER [84]. The quality of the models was assessed by Ramachandran plot analysis through PRO-CHECK [85]. The images were generated with PyMOL software (http:// pymol. org/) [86].

Genome sequencing of S. solidus Genome assembly and completeness
By combining short and long reads produced with Oxford Nanopore technology, we produced a genome assembly corresponding to 4944 scaffolds. These scaffolds summed to over 500 Mb (80% of the assembled genome). To test the completeness of our assembly, we mapped all the available transcripts of S. solidus [46] on our de novo assembly. We found that more than 99.6% (24,676/24,765) of transcripts mapped onto the new assembly. The remaining 89 transcripts that were not found were blasted (BLASTN) against the assembly dataset. A total of 46 transcripts had a hit on 28 scaffolds, but these hits were weak regarding the alignment length. This left 43 orphan transcripts that had no match in the assembly.
We also searched BUSCO groups to investigate the completeness of the genome assembly using a complementary approach. A total of 649 complete BUSCOS (66%) were found out of the 978 BUSCO groups in the Metazoa database, which is close to the proportions obtained from the assembly of other cestode genomes (e.g. 72.6% in the genome of Schistosoma japonicum [91]; 73.2% in the genome of Schistosoma haematobium [92]). The BUSCO duplication rate was 0.8%. Overall, these results suggest that the coding genome was well represented in the genome assembly. Finally, we identified a total of 56% of elements that appeared to be repeated in the new genome of S. solidus (Additional file 5), which is in accordance with previous studies that demonstrated that the genomes of most flatworms include large numbers of repetitive elements [93].

Number of putative genes
We first performed gene prediction using an ab initio approach, which gave a total of 21,780 ORFs. A second approach based on bam files from the transcript alignments of S. solidus [46] led to the identification of 30,103 ORFs. The two sets were merged, and we found that 9103 ORFs overlapped between the two sets, with a highly variable percent identity (ranging from 1 to 100%). Because of this variation in percent identity and in order to obtain exhaustive information, we decided to keep duplicate sequences for the rest of the analysis, which left a total of 51,883 ORF sequences identified. When comparing these predicted genes to the previously published transcriptome of S. solidus, we found that 77% of the transcripts were found in these predicted genes, and that 66% of the predicted genes were found in the transcriptome.

Identification of sequences specific to the new assembly
The 51,883 ORF sequences were aligned against a database of 43,058 protein sequences of S. solidus obtained from NCBI. We identified 19,853 sequences that had no BLAST match and 16,287 sequences that had no significant match (E-value < 1e−5). The total unmatched sequences were therefore 36,140. These unmatched sequences are potential protein-coding genes not represented in public databases of S. solidus and were therefore used as one of the reference databases in the LC-MS/MS analysis.

Annotation of the new expanded S. solidus genome
The 51,883 ORF sequences were annotated using domain information [70] and orthology assignment [71]. First, an important fraction of the predicted sequences (35,576 ORFs; 68.6%) did not have a match in any database, and therefore, functions could not be assigned to them. We refer to them as putative S. solidus-specific genes (Additional file 6). We found that 16,307 ORFs were successfully assigned a putative biological function (Additional file 6: 7796 ORFs had a match on both databases, 4591 ORFs had a match on the domain database only, and 3920 ORFs had a match on the orthology database only). Enrichment analyses on the new expanded S. solidus genome were conducted to identify a potential class of genes that could be overrepresented and that could be associated either with the S. solidus parasitic lifestyle or with the phenotypic changes in the fish host. Overall, enrichment analyses showed that the functions were related to environment sensing: 2.5% of ORFs are rd3 genes known to be expressed in photoreceptor cells [94]; cell division, growth, and development: 1.7% of ORFs encode for proteins involved in the movement of microtubules, and 1.2% of ORFs are atrophin-coding genes involved in development [95]; as well as cell physiology: 1% of ORFs encode for PARP proteins involved in various cell physiological processes [96].

Characterization of the proteome
Using mass spectrometry (LC-MS/MS), we detected 2290 proteins in samples of whole S. solidus tissues, among which 1467 proteins were detected in all worm samples (Additional file 7). Among these 2290 proteins, 246 proteins were only detected using the new genome of S. solidus as a reference database during the LC-MS/ MS analysis, with 113 proteins detected in all worms. The new genome sequence and annotation therefore provide a significant additional resource for S. solidus functional genomics. Most of the 2290 proteins were functionally annotated, with the exception of the 246 proteins previously mentioned, for which functions were further investigated using a three-step approach based on sequence, structure, and phylogenetic analyses (see below). Using this approach, we found that the 2290 proteins of S. solidus included 40 proteins encoded by 27 genes that did not have any sequence or domain similarities with other known species, and for which functions could not be described (i.e. 1.7% of putative S. solidus-specific proteins in the proteome, compared to 68.6% of putative S. solidus-specific genes in the new genome) (Additional file 8).
Enrichment analyses in terms of biological process, cellular component, and molecular function (GO terms) were performed for all the proteins detected in at least one of the worm samples, and also for the proteins detected in all worm samples. For proteins with several protein IDs, we kept only the first ID in order to prevent enrichment bias. We found that 14.9% of proteins detected in all worm samples were involved in protein metabolism processes, namely "translation" [GO:0006412] and "protein folding" [GO:0006457] (p < 0.001). These are typically highly expressed proteins, which explains their overrepresentation here [93]. Also, we found that 10.2% of proteins detected in all worm samples referred to mechanisms important in the functioning and the regulation of the cell cycle: "microtubulebased process" [GO:0007017], with the microtubules having major roles in cell division and growth, and "tricarboxylic acid cycle" [GO:0006099], the latter referring to abundant metabolic enzymes that are involved in respiration by producing ATP (p < 0.001) (Fig. 2a). We also found that 14.4% of proteins detected in all worm samples were proteins with GTP binding [GO:0005525] and GTPase activity [GO:0003924] (p < 0.001) (Fig. 2b). Consistent with all the functions described above, we found that 23.2% of proteins detected in all worm samples were localized in the cytoplasm (p < 0.001).
The prediction of the functions of the 246 proteins that were detected using the new genome as a reference database was further investigated with sequence and structure analyses (Additional file 7). Their functions were in accordance with the previous enrichment results, as we found for example proteins involved in microtubules (protein ID: g20896.t1) or with ATPase activity (protein ID: g10622.t1). Additionally, we found that six proteins were peptidases or proteases with conserved active sites, suggesting that they are functional (  Fig. 2 The proteome and secretome of S. solidus include proteins involved in protein metabolism/cell growth/energy intake. Results that were significant (p < 0.001) for the enrichment analyses performed in terms of biological process (a) and molecular function (b). In each case, analysis was performed for the proteins detected in all worm (outer chart) and secretome (inner chart) samples hydrolase like (SsFAAH-like). FAAH enzymes degrade signalling lipids of the endocannabinoid class [97]. Additionally, 141 proteins were assigned both to the worm and to the fish host during LC-MS/MS analysis (Additional file 9). As it was not possible to directly determine if the 141 proteins were produced by the worm or the fish, we searched if these proteins could be core proteins, i.e. proteins conserved in all eukaryotes and whose functions are well characterized [98]. Using BLASTP and Pfam tools, we found that the 141 proteins had sequence similarities with 248 previously reported eukaryotic core proteins [98] (mean Expectvalue 2.45e−11) and that they were composed of conserved domains. We therefore concluded that these proteins were expressed both in the worm and in the fish because of their fundamental functions in the cell, but that their peptide similarities did not allow us to determine their origin. Enrichment analysis performed on these 141 proteins confirmed that they were involved in cell physiology and in energy production as found for the whole proteome.

Characterization of the secretome
A total of 1568 proteins were detected in the secretome samples (Additional file 7). The numbers ranged between 781 and 1183 proteins depending on the sample, with 459 proteins detected in all secretome samples. As expected, the secretome of S. solidus was mostly a subset of the worm proteome, both in terms of protein number and functions, with a few exceptions described below. The whole protein content of the secretome of a given worm represented up to 59% of the whole protein content of the proteome of that same worm. In total, 1538 unique proteins were shared between at least one secretome sample and one proteome sample, among which 385 proteins were detected in all secretomes and proteomes studied, and 74 proteins were detected in all secretomes and in at least one proteome. All the proteins that were shared between secretome and proteome samples had a functional annotation, except for 36 proteins that were previously detected in the proteome as S. solidus-specific and that were also secreted (2.3% of putative S. solidusspecific proteins in the secretome, compared to 68.6% of putative S. solidus-specific genes in the new genome and 1.7% of putative S. solidus-specific proteins in the proteome) (Additional file 8).
We found that the secretome was composed of proteins enriched for biological processes, cellular components, and molecular functions that were similar to the proteome. For the proteins detected in all secretomes, almost half of the proteins participated in the regulation of cell division and energy production. These proteins were enriched in functions such as the "microtubulebased process" [GO:0007017] involved in the regulation of cell division, and "glycolytic process" [GO:0006096] and "gluconeogenesis" [GO:0006094], these two processes being an important source of energy production (p < 0.001 hypergeometric test corrected with Bonferroni method) (Fig. 2a). Proteins were also enriched in domains with GTP binding [GO:0005525] and GTPase activity [GO:0003924] (Fig. 2b), as previously reported in the proteome (p < 0.001). Furthermore, 37.9% of the proteins detected in all the secretome samples were predicted to be localized in the cytoplasm (p < 0.001), similar to proteins in the proteome. Notably, we found that a significant number of proteins (6.1% of the proteins detected in all secretomes p < 0.001) were reported to be specifically localized into the extracellular space, an enrichment that was specific to the secretome.

Proteins unique to the secretome
We found eight proteins that were detected in all secretome samples, but in none of the proteomes (Table 3). Among them, three proteins had fibronectin type-III domains. The first (protein ID: A0A0X3PH69) was a "Neogenin", which is a protein involved in neural development with two fibronectin type-III domains. The second (protein ID: A0A0X3Q1B7; A0A0X3PKA1; A0A0X3Q8R6) was a receptor-type tyrosine-protein phosphatase eta that included three fibronectin type-III domains. The third (protein ID: A0A0V0JBL5) included two fibronectin type-III domains, but no additional information on the function of this protein was available. Furthermore, we detected an uncharacterized protein with a predicted molecular function corresponding to "neurotransmitter: sodium symporter activity (NSS)", which is involved in transmembrane transport (protein ID: A0A183S8K9; A0A0X3PDV9; A0A0X3PNC8; A0A0V0J682; A0A183T7R5), and a protein potentially acting at the cell membrane as a phospholipid scramblase (PLSCR) (protein ID: A0A0X3P711; A0A183SGM7). Lastly, three other uncharacterized proteins did not have information on function using the UniProt database, BLAST tools, and protein domain identification. It seems that these three proteins are specific to S. solidus, with the last protein (protein ID: A0A0X3Q756) having a secretory signal peptide. Twenty-two proteins were detected in at least one of the five secretome samples, but in none of the proteomes (Table 4). Three proteins were composed of fibronectin type-III domains. The first (protein ID: A0A0X3NX35) detected in four out of five secretomes was described as a receptor-type tyrosine-protein phosphatase H. The second (protein ID: A0A183TLI3; A0A0X3PWG6; A0A0X3PW04; A0A0V0J316; A0A0X3PN23; A0A0X-3PJY8), which was detected in three secretomes, was either a tenascin (an extracellular matrix glycoprotein) or a receptor-type tyrosine-protein phosphatase F. The last protein of this type (protein ID: A0A0V0JA38), also detected in three secretomes, was, according to the BLAST result, a collagen-like protein, which is an important component of cuticle.
Two proteins appeared to be involved in immunity processes. The first (protein ID: A0A183TPG4;A0A0 X3P3D7;A0A0X3PTB8), which was detected in three secretomes, was uncharacterized and had a cystatin domain which may be involved in immunomodulatory functions. The second (protein ID: A0A0V0JBV1; A0A0V0J795; A0A0X3NVP3; A0A0X3P0M4), which was detected in the secretome of the largest worm only (worm 4), was according to the BLAST results an antigen similar to the diagnostic antigen gp50 commonly used to detect parasitic diseases.
Two proteins were associated with transport functions. The first (protein ID: A0A183TIR8; A0A0X3PCX3; A0A183TT84; A0A0X3PT59) was detected in four out of five secretomes and was a sodium/glucose cotransporter involved in glucose homeostasis. The second (protein ID: A0A0X3NT74), which was only detected in the second largest worm (worm 2), was an intraflagellar transport protein required for ciliogenesis.
One protein (protein ID: A0A183TDP7), detected in four secretomes, had a knottin fold similar to a domain found in the Schistosoma parasitic trematodes, but for which the function is not characterized. It is a cysteinerich protein and interestingly, BLASTP search against the UniProt database revealed that it is only found in Platyhelminthes, and amino acid multiple sequence alignment showed that it presents a new motif, C-x[6]-C-x(7)-CCx(4)-C-x(9)-C-x[2]-C-x[6]-C-x(5)-CC-x(3)-C-x(4)-C. This type of protein is resistant to proteases.
Nine proteins had functions which could not be clearly determined by BLAST in UniProt and nonredundant protein sequences (GenBank) databases. Five other proteins were only detected using the new genome of S. solidus as a database reference. We further investigated the functions of these last 14 proteins with sequence, structure, and phylogenetic analysis: nine proteins were specific to S. solidus. Among them, two proteins (protein ID: A0A0V0J2U1; protein ID: A0A0X3PIM2) had secretion signal peptides, and one protein (protein ID: A0A0X3PXG6) had a transmembrane helix (TMH) domain found in transmembrane proteins. Furthermore, one protein was identified as a peptidase M28B (glutamate carboxypeptidase 2) Table 3 Proteins that are excreted/secreted by S. solidus, detected in five secretomes and not in proteomes Protein IDs were taken from UniProt. When several protein IDs were assigned to one protein, these protein IDs corresponded to isoforms with identical functions. For each protein, functional annotation was first retrieved from searches with UniProt and BLAST tools. Complementary analyses based on sequence, structure, and phylogeny were used to obtain information

Discussion
Understanding the molecular mechanisms behind behavioural changes in a vertebrate host infected by a non-cerebral parasite is a fascinating challenge. If these changes are the result of parasite manipulation of its host, one possible mechanism hinges on molecules secreted by the parasite (i.e. manipulation factors) that impact the host's physiological, immunological, and central nervous systems, and ultimately the behaviour of the host. Here, we describe for the first time the protein component of the secretome of a parasite commonly referred as manipulative, using mass spectrometry (LC-MS/MS), with the objective to identify potential manipulation factors that could explain host behavioural changes. As expected, we found that the proteins that are excreted/secreted by S. solidus are mostly a subset of the proteins being expressed in the whole worm and that both included proteases. We also found that 30 secretome proteins were not detected in the proteome and were involved in neural, immune, and cell communication functions, therefore having the potential to interfere with the host's physiological systems and behaviour. Finally, we highlighted that the secretome of S. solidus included S. solidus-specific proteins that could play important roles in the tight interaction of the parasite with its fish host. All together, these proteins represent promising candidates to explain physiological and behavioural changes in the stickleback host.

The secretome of S. solidus is mostly a subset of its proteome with similar functions
The global protein composition of the proteomes and the secretomes of S. solidus highlights functions that are crucial for its parasitic lifestyle and for its interactions with the external environment. As it is generally reported [52], we found that the proteins that are excreted/secreted by S. solidus (i.e. the secretome) are mostly a subset of the proteins being expressed in the whole worm (i.e. the proteome), both in terms of protein number and function. We found that both the proteome and the secretome of S. solidus at the infective stage (i.e. when behavioural changes appear in the fish host) were enriched in proteins involved in cell division, which was also found in the functional analysis of the genome we sequenced, as well as energy production. Concerning cell growth, our results may seem counterintuitive because a previous transcriptome analysis revealed that the non-infective stage of S. solidus shows higher expression levels of genes involved in growth and cell regulatory functions compared to the infective stage [46]. However, we did not analyse the proteome and the secretome of worms at the non-infective stage, such that we can only quantify that these processes are enriched at the infective stage but not their relative importance compared to other life stages of the parasite. We therefore only speculate that both the non-infective and the infective stages of S. solidus could rely on biological processes involving cell growth, but production levels would be much higher at the non-infective stage, considering the fact that growth occurs predominantly in the first 12 weeks after installation in the fish host [99]. Concerning energy production, the results are also surprising, as the worm was empirically described to use its glycogen reserve (its primary source of energy) mainly when it reaches its final avian host [100]. However, transcriptomic analyses demonstrated that glycogen metabolism and energy production in S. solidus are complex processes [46]. Specifically, six steps of the glycolysis cycle are highly expressed at the infective stage [46]. Furthermore, it is important to keep in mind that for tissue and secretome samples collection, worms were placed in a solution of PBS (pH 7.4) in the dark for 2 h. While this method is commonly used to collect parasitic secretomes [73], we cannot dismiss the possibility that the environmental change (from the body cavity of the stickleback to the PBS tube) encountered by S. solidus at this step could have influenced the levels of proteins detected in the proteomes and secretomes. One can speculate that this environmental stress could have triggered cell growth and energy intake in the worm (for example, to counteract other metabolic effects), explaining the discrepancies between the proteomics data we obtained and the previous transcriptomic work [46]. To confirm our results, future work could replace the PBS solution by a medium that would better mimic the fish abdominal cavity, for example, by a medium already used in vitro for breeding [32]. The global analysis of the proteome and of the secretome of S. solidus therefore confirms the importance of energy use for the worm, even before it reaches the bird. We detected both in the proteome and the secretome of S. solidus proteins with GTP binding and GTPase activity. GTP-binding proteins also called G-proteins are known to regulate a variety of biological processes such as mediating signals by hormones and light, gene expression, cytoskeletal and microtubule organization, or vesicle trafficking [101][102][103][104]. In parasites, GTPases have been demonstrated to have important roles in the secretion of virulence factors [103,104]. For instance, in Toxoplasma gondii, which is a parasite known to induce behavioural changes in rodents [105], Rab GTPases regulate the secretion of proteins essential to invade host cells, and the modification of their expression results in aberrant transport of proteins [106].
The use of the new genome of S. solidus as a reference database during mass spectrometry analysis allowed us to detect six peptidases or proteases in the proteome of S. solidus, and one peptidase (peptidase M28B glutamate carboxypeptidase 2) detected only in the secretome fraction, with proteases and peptidases being typically described as virulence factors for many parasites (Table 1). In S. solidus, proteases and peptidases may have important roles in weakening the fish immune response at the infective stage when the worm is ready to pass into its avian host to reproduce [32]. Injecting these proteases and peptidases, alone or in combination, into isolated head kidney samples from non-infected fish as in Scharsack et al. [32] would be necessary to confirm their potential role in disrupting the host immunity. Furthermore, this shows that the new genome presented here is a valuable tool to identify proteins that are critical for the parasitic lifestyle. We also found an SsFAAH-like enzyme that could potentially degrade signalling lipids of the endocannabinoid class. Endocannabinoids have previously been reported as an important player in hostparasite interactions by promoting the activation of the immune response in the host [97,107]. This enzyme was not detected in the secretome of S. solidus.

Neuronal and immune functions
Transcriptomic analysis demonstrated that when S. solidus reaches the infective stage, genes involved in neural pathways and sensory perception are expressed by the worm at higher levels [46]. Thus, we expected the secretome at the infective stage of S. solidus to be enriched with proteins involved in such neural functions (as described in other parasitic systems, reviewed in Table 1). Furthermore, behavioural changes in the stickleback infected by S. solidus generally appear in concomitance with the activation of the immune response of the fish [32]. Therefore, we expected the secretome of S. solidus at the infective stage to include proteins involved in immunity (as reviewed in Table 1). We found that three proteins were playing potential roles in neural and immune functions. The first protein, which was detected in four secretomes but in none of the proteomes, had a knottin fold called UPF0506 composed of cysteines and generally found in Schistosoma parasites. In Schistosoma, the function of the proteins with such knottin fold is not defined. However, peptides with knottin domains (i.e. knottins) were described in venoms from various animals. For venomous animals, knottins are neurotoxins having high specificity towards receptors in the nervous system of their prey or aggressor [108]. Furthermore, it is a cysteine-rich protein, and in parasites, cysteinerich proteins play a role in invasion [109] and modulation [110] of the immune system. Therefore, this protein could be a promising manipulation factor if it could act as a neurotoxin in the brain of infected sticklebacks. The second protein, which was detected in three secretomes but in none of the proteomes, had a cystatin domain which may be involved in immunomodulatory functions. It was demonstrated in parasitic nematodes that cystastins are important secreted molecules that help parasites to evade immunity of the host [111]. Nematode cystatins inhibit host proteases involved in antigen processing and presentation, leading to a reduction of the host immune response [112]. This secreted protein could thereby explain in part why the immune response is only activated late in infected sticklebacks, which needs to be further studied. The third protein appeared to be an antigen similar to the diagnostic antigen gp50. This diagnostic antigen is used to detect parasitic infection, for example, by Taenia solium [113]. However, the antigen was detected only in the secretome from the biggest worm of our study and not in its proteome. Sampling over a longer time frame could alleviate this type of discrepancy between samples. One protein, which was detected in all secretomes but in none of the proteomes, had a "neurotransmitter: sodium symporter activity (NSS)" annotation. In humans, it is a membrane protein involved in the termination of synaptic transmission and the recycling of neurotransmitters at the membrane [114]. How this membrane protein could be secreted and act as a manipulation factor in the secretome is unclear. Therefore, we are cautious in interpreting the role of this protein as a manipulation factor, as it could be solely a waste product from the membrane of the parasite.

Cell communication
Schistocephalus solidus is located in the abdominal cavity of the threespine stickleback [27]. As the worm is not in direct contact with its host brain, we expected to detect proteins involved in cell communication or cellcell signalling. During annotation of the new genome, 2.5% of genes were related to environmental sensing functions. Furthermore, we found that seven proteins detected only in the secretome fraction were characterized by fibronectin type-III (FNIII) domains. FNIII domains are widely found in animal proteins and are involved in cell-cell interactions [115]. The first protein with fibronectin type-III domains found in all secretomes but in none of the proteomes was identified as a neogenin. In addition to playing roles in cell-cell adhesion, neogenins are involved in neural development in humans [116]. Three additional proteins that we identified with fibronectin type-III domains were described as receptortype tyrosine-protein phosphatases (type eta, detected in all secretomes but in none of the proteomes; type H, detected in four secretomes but in none of the proteomes; type F, detected in three secretomes but in none of the proteomes). Receptor-type tyrosine-protein phosphatases are involved in cell-cell communication [117]. It was previously shown that a baculovirus secretes a protein tyrosine phosphatase, which acts on the neural system of its host, the silkworm Bombyx mori, and enhances its locomotory activity, so that it ultimately increases the virus dispersal [19] ( Table 1). The set of phosphatases identified here are located in the membrane, such that they are predicted to have specific biological functions that differ from those of the protein tyrosine phosphatase found in the baculovirus. However, their abundant representation in the secretome is intriguing, and the importance of the process of phosphorylation represents a future avenue of study in the context of behavioural changes in the fish host, such as increased exploration [34]. For the next two proteins identified with fibronectin type-III domains, the first protein was found in all secretomes but in none of the proteomes and had no clear function (protein ID: A0A0V0JBL5), as for the second protein that was detected in four secretomes but in none of the proteomes (protein ID: g17644.t1). The last protein appeared to be a collagen-like protein found in three secretomes. In Caenorhabditis elegans, the cuticle is an extracellular matrix made of small collagen-like proteins [118]. Thus, the role of this protein in cell communication is uncertain, and it may rather be a waste product excreted passively from the cuticle of S. solidus. In summary, fibronectin type-III proteins secreted by S. solidus appear to be good candidates for manipulation factors because of their roles in cell-cell signalling, but also in potential neural functions.
Additionally, we expected to detect proteins with transport functions in the secretome of S. solidus to mediate the communication between the worm localized in the host abdominal cavity and the host brain. We detected three proteins involved in transport or related functions at the membrane, including a protein that was detected in all secretomes but in none of the proteomes, a phospholipid scramblase. It is a transmembrane protein that is known in humans to bind to the 5'-promoter region of the inositol 1,4,5-triphosphate receptor type 1 gene (IP3R1) so that it enhances expression of the receptor [119]. Since these proteins detected here could be simply waste products coming from the membranes of the parasite, we thus have to be careful when speculating about the roles of these proteins, and functional analyses will be required.

The secretome includes S. solidus-specific proteins that could play important roles in parasite-host interactions
During annotation of the new genome, we found that 68.6% of the predicted genes did not have any sequences or domain similarities with other known species that would enable accurate annotation, therefore representing putative S. solidus-specific genes. Part of these genes are effectively translated, as 1.7% of the proteome and 2.3% of the secretome included putative proteins specific to the worm. Specifically, we detected a total of 12 proteins only in the secretome fraction for which functions could not be clearly determined, but that were specific to S. solidus. Three proteins were detected to be secreted signal peptides (protein ID: A0A0X3Q756; protein ID: A0A0V0J2U1; protein ID: A0A0X3PIM2). We found that the proteins that are specifically enriched in the secretome are more likely to have secreted signal peptides (10% of the proteins detected only in the secretome fraction had secreted signal peptides, while 0.17% of the proteins detected in the proteome had such signals). One protein had a transmembrane helix (TMH) domain found in transmembrane proteins (protein ID: A0A0X3PXG6). Proteins with TMH domains are used to diagnose parasitic infection, as they are highly specific to the parasite of interest [120,121]. Interestingly, analyses conducted with the transcriptome of S. solidus previously suggested that 19% of all the protein-coding genes could be S. solidus-specific [45]. In this study, producing a novel genome sequence and assembly of S. solidus led us to increase this estimation by three times. S. solidusspecific secreted proteins represent promising candidates to explain physiological and behavioural changes in the stickleback host. The threespine stickleback is an obligatory host of S. solidus [122]. Because of the potential coevolution between the worm and the fish [27], S. solidus may have developed highly specific molecular mechanisms targeting the threespine stickleback physiological machinery to insure it can establish and grow in this fish only, until it is ready to pass into its final avian host. The S. solidus-specific proteins found here are therefore likely to be part of this unique set of molecules crucial for the survival of the worm, and we can assign them this ecological annotation [123]. It will be very interesting in the future to produce recombinant S. solidus-specific proteins and test their effects on the phenotype of the threespine stickleback by functional analysis. These proteins could also serve to obtain structural information by X-ray crystallography or NMR spectroscopy, which could help us to discover their function. Finally, studies of how these proteins evolve between Schistocephalus populations and between species of Schistocephalus infecting different fish species [124] will shed additional light on the specificity and potential function of these proteins.

Conclusions
The secretome of S. solidus appears to be an important component of the molecular interaction between the parasite and its threespine stickleback host. In accordance with our predictions, we detected in the secretome of S. solidus at the infective stage (i.e. when behavioural changes appear in the stickleback) proteases, proteins involved in neural and immune functions, as well as proteins involved in cell communication. We also highlighted in the secretome the presence of S. solidus-specific proteins. In the future, comparative studies could be conducted to validate that the proteins detected in the secretome of S. solidus act as manipulation factors in the behaviour of its fish host: at the organism level, the analysis of the secretome of the worm at the non-infective stage will confirm if the putative manipulative proteins reported in this study are effectively detected solely at the infective stage (or at highest levels) when the behavioural changes appear in the host. It would also be of special interest to study the secretome of S. solidus at the stage when it is able to reproduce, i.e. in the bird host. At the species level, comparing the secretome of S. solidus with the secretome of a closely related species Schistocephalus pungitii will be of great interest, as S. pungitii does not have known effects on the behaviour of the nine-spine stickleback, its specific intermediate host species [125]. Finally, functional studies testing effects of presence in non-infected sticklebacks of interesting proteins identified in the secretome, such as receptor-type tyrosineprotein phosphatases or phospholipid scramblase, would allow us to better understand the contribution of these proteins in the behavioural changes. To conclude, we hope that the genomic and proteomic resources we provide will help other researchers to investigate general questions on host-parasite interactions in nature.