Background

Food-borne trematode (FBT) infections are important neglected tropical diseases (NTDs) with a global public health impact estimated at more than 665 thousand disability-adjusted life years (DALYs); paragonimiasis is arguably the most important of these because it accounts for nearly 30 % of the FBT-related DALYs [1]. Approximately 20 million people already have a Paragonimus infection, and almost 300 million people are at risk of becoming infected [2, 3].

More than 50 species in the genus Paragonimus have been described, although several could be considered synonymous [4]. Fifteen species are known to infect humans, but the P. westermani and P. skrjabini species complexes are responsible for the bulk of disease in Asia, particularly in the People’s Republic of China, which has the heaviest disease burden among 48 endemic countries [3].

The life-cycle of Paragonimus flukes involves complex interactions with three separate hosts [3]. Embryonated eggs expelled in the sputum or feces hatch in freshwater, releasing larvae that undergo rounds of growth and asexual reproduction in the first intermediate host, an aquatic snail. The snails, in turn, release larvae that develop into metacercariae in crustaceans. When infected crustaceans are ingested by a permissive host (typically small carnivores such as canids, felids, murids, mustelids, viverrids, etc.), metacercariae migrate out of the digestive tract and into the lung, where they mature to long-lived, hermaphroditic, sexually reproducing adults within pulmonary cysts. In contrast, metacercariae ingested by a non-permissive often fail to find the lung. They remain in an immature state and migrate through abnormal tissues including the central nervous system (CNS). Paragonimus skrjabini, for example, is poorly adapted to humans and often causes these ectopic infections [3].

Paragonimiasis is commonly diagnosed by detecting parasite eggs in stool or sputum. Unfortunately, the time interval between infection and oviposition is typically 65–90 days [3], and migrating parasites are capable of causing disease much sooner than this [1]. Migration of worms through the abdominal cavity can cause diarrhea, abdominal pain, fever and hives. Parasites in the lung trigger asthma- or tuberculosis-like symptoms with including cough, fever, pleural effusion, chest pain and bloody sputum. Ectopic infections in the CNS can lead to headache, visual loss, or death if left untreated [1]. Paragonimiasis is easily treated with oral praziquantel. However, diagnosis and treatment are often delayed, because of the non-specific nature of the symptoms and the lack of sensitive and reliable diagnostic methods [5].

Apart from widely used phylogenetic markers, Asian Paragonimus species are very poorly represented in pubic sequence repositories. In the year 2015, there were only 456 protein sequences from the genus Paragonimus in NCBI’s non-redundant protein database (NR). This represents a significant hindrance to the biological research that will be needed to promote the development of novel methods for diagnosis, treatment and global control of paragonimiasis. In order to address this need, we have sequenced and characterized the transcriptomes of P. westermani and P. skrjabini adult worms. Transcriptome sequencing is a well-established, efficient, and cost-effective method of gene discovery that has been used to characterize the expressed genes of trematodes and other parasites [6–8].

Thus, our study has provided insights into the biology of two Paragonimus species along with a wealth of novel sequence data that could be explored to test specific hypotheses relating to Paragonimus and other FBTs.

Methods

Parasite material

Freshwater crab intermediate hosts were collected to obtain parasite metacercariae. Crabs of the genus Isolaptamon were collected from Liuyang county (now called Baisha county), Hunan Province, China, a region specifically endemic to P. westermani [9]. Likewise, Sinopotamon denticulatum were collected from Changan county of Shanxi Province, China, a region specifically endemic to P. skrjabini [10]. Metacercariae were isolated from crab tissue as previously described [11]. The shells of the crabs were removed and the soft tissues were processed in 1× phosphate-buffered saline with a meat grinder. The homogenized meat was allowed to settle, and the supernatant was discarded. The sediment was rinsed several times in water, and metacercariae were collected under a dissection microscope. Species identity was confirmed by morphological examination of metacercariae and later by examination of adult parasites [12–15].

Dogs obtained from non-endemic areas and clear of existing infections were inoculated orally with 200–300 P. westermani or P. skrjabini metacercariae. Adult worms were harvested from the lungs 100 days post-infection, washed thoroughly in diethylpyrocarbonate-treated water, frozen in liquid nitrogen, and stored at -80 °C prior to use.

RNA isolation and sequencing

A total of 5 adult P. westermani and 5 adult P. skrjabini were homogenized in 1 ml TRIzol reagent with microcentrifuge pestle, and total RNA was purified from the homogenate using a TRIzol Plus RNA Purification Kit manufacturer’s recommended protocol (Thermo Fisher Scientific, Waltham, MA, USA) and DNase-treated. Samples had very prominent 28S peaks and very small 18S peaks, with RIN values and DV200 values of 8.3 and 71 (P. westermani, concentration 677 ng/μl) and 7.7 and 72 (P. skrjabini, concentration 562 ng/μl), respectively (Additional file 1: Figure S1). Sequencing libraries were prepared from 2 μg total RNA using Illumina's TruSeq Stranded mRNA Library Preparation Kit according to the manufacturer’s protocol and sequenced on the Illumina HiSeq2000 platform (Illumina, San Diego, CA, USA). Raw reads (100 bp in length) were deposited in the NCBI sequence read archive under BioProject ID PRJNA219632 for P. westermani and PRJNA301597 for P. skrjabini.

RNA-Seq read processing and assembly

Raw reads were subjected to stringent quality control and contaminant filtering as previously described [16]. Briefly, reads were trimmed to remove low quality regions, and filtered based on read length, sequence complexity, and similarity to known or suspected contaminants, including ribosomal RNA [17, 18], bacteria [19], Homo sapiens (GenBank version hs37) and Canis familiaris (GenBank version 3.1). Remaining high-quality, contaminant-free read sets were down-sampled by digital read normalization using khmer (k = 20) [20]. Reads selected in the down-sampling and their mates were assembled using the Trinity de novo RNA-Seq assembler using default parameters [21]. Scripts included in the Trinity software package were used to map the complete, cleaned read set to the assembled transcripts and filter transcripts less than 1 transcript per million reads mapped and less than 1 % of the per unigene expression level [21]. Assembly fragmentation was calculated with respect to Clonorchis sinensis coding sequences (WormBase ParaSite BioProject PRJDA72781) using in-house scripts and is reported as the percentage of reference genes matched to multiple, non-overlapping transcript BLAST hits.

Transcript expression analyses

The complete, cleaned read sets were mapped to the corresponding filtered, high-quality transcript assemblies, and fragments per kilobase of exon per million mapped fragments (FPKM) were calculated for each unigene according to an RNA-Seq by expectation-maximization (RSEM) protocol using scripts included in the Trinity software package [21]. Unigenes were ranked according to abundance based on FPKM values. Fold changes were calculated for the corresponding unigenes from the two assemblies. The average fold change plus or minus 1.96 times the standard deviation (corresponding to the top 5th percentile of up-/downregulation) was used as a cut-off to select unigenes that were differentially expressed between the two species.

Protein prediction and functional annotation

Protein sequences were predicted from transcripts using Prot4EST [22] based, in part, on results from BLAST searches against the NCBI non-redundant protein database (NR, downloaded on 15 April 2014) and databases of ribosomal [17, 18] and mitochondrial genes (downloaded from GenBank on 26 July 2013).

Protein translations were compared to known proteins in NR (downloaded on 15 June 2015), Clonorchis sinensis (WormBase ParaSite BioProject PRJDA72781), Opisthorchis viverrini [23], Fasciola hepatica [24] and Paragonimus kellicotti [16] protein sequences by BLASTP, and results were parsed to consider only non-overlapping top hits with e-value ≥ 1e-05. Sequences from Paragonimus species were excluded from NR prior to BLAST searches in order to facilitate identification of genus- and species-specific transcripts. The longest predicted protein isoform of each assembly unigene was also subjected to a reciprocal best BLAST match between the P. skrjabini and P. westermani transcripts with an e-value cut-off of 1e-05.

Predicted proteins were matched to conserved domains (InterPro) and gene ontology (GO) terms using InterProScan [25–27]. Associations with biological pathways (KEGG orthologous groups, pathways and pathway modules) were determined by KEGGscan [28, 29] using version 70 of the KEGG database. KEGG module completion was determined as previously described [30]. Putative proteases and protease inhibitors were identified and classified by comparison with the MEROPS database [31]. Classical secretion signals found within the first 70 N-terminal amino acids and transmembrane domains were predicted with Phobius [32]. All assembled transcripts, predicted proteins, and associated functional annotations are available at Trematode.net [33].

Identification of “host mimic” proteins

The longest isoform of each assembly unigene was compared to proteins from Homo sapiens (NCBI hs38) and the closest sequenced free-living relative, Schmidtea mediterranea (WormBase ParaSite Bioproject PRJNA12585), by BLASTP. Deduced Paragonimus proteins were considered putative “host mimics” when they shared at least 70 % sequence identity over at least 50 % of the length with the human ortholog but less than 50 % identity (if any) with the S. mediterranea ortholog.

Functional enrichment of gene ontology (GO) terms

Functional enrichment of GO terms was calculated using FUNC with a P-value cut-off of 0.01 [34]. In all cases, the target list was comprised of the longest transcript of each unigene associated with the feature of interest and the background list was comprised of the target list plus the longest transcript from each remaining unigene.

Results and discussion

Transcriptome sequencing, assembly and annotation

The adult transcriptomes of P. westermani and P. skrjabini were sequenced, assembled de novo, and filtered to consider only high-confidence transcript sequences (Table 1). In each case, related transcripts thought to result from alternative splicing of the same gene were clustered into “unigenes”. A total of 27,842 transcripts from 21,586 unigenes were generated from P. westermani while 35,312 transcripts from 25,825 unigenes were generated from P. skrjabini. Unigenes from the two species had similar length distribution patterns (Fig. 1). We expect these species to encode a gene complement similar in size to those of other FBTs: 13,634 for C. sinensis [35], 16,379 for O. viverrini [23] and 15,740 for F. hepatica [24]. In an ideal assembly, the number of unigenes would equal the number of genes expressed genes in the life-cycle stage or condition studied. However, de novo short read assemblies tend to be fragmented, and this inflates unigene counts. Fragmentation, reported as the percentage of reference genes matched to non-overlapping transcript BLAST hits, was estimated at 24.3 % for P. westermani and 26.7 % for P. skrjabini with respect to the protein coding sequences of C. sinensis. For clarification, this indicates that 24.3 % of all C. sinensis genes are associated with multiple, non-overlapping P. westermani transcripts.

Table 1 Sequencing, assembly and annotation of the transcriptomes of adult P. westermani and P. skrjabini
Full size table
Fig. 1
figure 1

Unigene length distribution for Paragonimus westermani (a) and Paragonimus skrjabini (b)

Full size image

A total of 26,431 and 32,796 unique protein translations were generated from P. westermani and P. skrjabini respectively, and these were annotated based on similarity to sequences in various publicly available databases (Table 1). Complete annotations are provided in Additional file 2: Table S1 and Additional file 3: Table S2. Altogether, functional information (e.g. BLAST matches, structural domains, functional classification, etc.) was deduced for a majority of unigenes, 79.3 % and 80.0 % for P. westermani and P. skrjabini, respectively.

Sequence conservation with relevant trematode species

Due to the sparse representation of Paragonimus sequences in public sequence repositories, only a small fraction of our predicted proteins shared highest sequence similarity with Paragonimus sequences in NR (125 transcripts from 86 P. westermani unigenes and 151 transcripts from 88 P. skrjabini unigenes); a majority of these also had close matches to non-Paragonimus sequences. Predicted proteins from 69.8 % and 60.6 % of P. westermani and P. skrjabini unigenes, respectively, had top matches to non-Paragonimus proteins in NR (Additional file 2: Table S1 and Additional file 3: Table S2) due to the underrepresentation of Paragonimus spp. references in NR. Top hits were mostly to other food-borne trematodes, particularly C. sinensis and O. viverrini. Some 1,217 of the 6,513 P. westermani and 10,171 of the P. skrjabini unigenes with no significant match to non-Paragonimus proteins in NR were homologous in both species (i.e. conserved hypothetical unigenes, Fig. 2). This strengthens the notion that they are indeed valid (not caused by assembly errors), Paragonimus-specific transcripts.

Fig. 2
figure 2

Paragonimus-specific proteins from Paragonimus westermani and Paragonimus skrjabini. Predicted proteins from 6,513 P. westermani unigenes and 10,171 P. skrjabini unigenes found no significant match to non-Paragonimus sequences in NR. Of the unigenes with no BLAST match in NR, 1,217 from each assembly were matched to a non-hit unigene in the other assembly

Full size image

Comparisons to other trematode species at the primary sequence level indicated that deduced proteins from P. westermani and P. skrjabini share higher sequence identity with proteins from P. kellicotti (the only Paragonimus species with an available adult transcriptome) compared to other FBTs (Table 2). Paragonimus westermani and P. skrjabini may share slightly higher sequence identity with C. sinensis as compared to O. viverrini and F. hepatica; however, this result may be biased by the quality and completeness of the genome assemblies and gene models included in the analysis, as phylogenetic analyses based on mitochondrial markers have previously placed Paragonimus alongside F. hepatica rather than the carcinogenic liver flukes [36, 37].

Table 2 BLASTP comparisons of P. westermani and P. skrjabini proteins with selected trematode species. The number of transcripts/unigenes with BLASTP match (e-value < 1e-05) to subject proteins is indicated. Average percent identity was calculated based on the top hit to the longest isoform of each unigene
Full size table

Metabolic potential of Paragonimus westermani and P. skrjabini

Translated proteins were matched to KEGG orthologous groups and their parent unigenes were binned into broad functional categories (Table 3). The most abundantly populated categories from both assemblies were “signal transduction”, “translation” and “protein folding, sorting, and processing”. Most of the InterPro domains and KEGG orthologous groups that were represented in the adult transcriptomes of P. westermani and P. skrjabini were also represented in the genomes of other food-borne trematodes (Fig. 3). The 1,989 conserved protein domains and 1,419 conserved KOs provide a catalog of functions involved in core biological processes common to all sequenced FBTs. Paragonimus westermani and P. skrjabini shared more InterPro domains with the genome of F. hepatica as compared to the genome of C. sinensis. Some 145 InterPro domains and 195 KEGG orthologous groups were represented in the transcriptome assemblies of both Paragonimus species but absent from the draft genomes of the other two flukes. These Paragonimus conserved/specific KEGG orthologous groups were involved in 28 unique modules, all of which were sparsely populated (Additional file 4: Table S3); therefore, it is difficult to comment on metabolic differences between Paragonimus and other FBTs based solely on the transcriptomes.

Table 3 KEGG categorization of assembled unigenes
Full size table
Fig. 3
figure 3

Distribution of InterPro domains and KEGG orthologous groups among selected food-borne trematodes. InterPro protein domains and KEGG orthologous groups were assigned to proteins from the complete genomes of Clonorchis sinensis and Fasciola hepatica and to proteins predicted from the Paragonimus westermani and Paragonimus skrjabini transcriptome assemblies, and intersections were determined. Abbreviations: Pw, P. westermani; Cs, C. sinensis; Fh, F. hepatica; Ps, P. skrjabini

Full size image

The metabolic potential of the two species was assessed at the level of KEGG pathway modules. A KEGG module is considered to be complete when the transcriptome includes the full complement of enzymes (assessed at the level of KO’s) necessary to convert the initial substrate to the final product. Of 95 helminth-relevant KEGG modules [30], 35 were complete in both P. westermani and P. skrjabini. A total of 30 complete modules are shared between the two, with five uniquely complete in each species. However, the difference between the complete modules in one species and the incomplete modules in the other is at most two KO’s, suggesting high functional conservation among the two species.

Other FBTs are known to undergo transitions in energy metabolism over the course of the life-cycle, shifting from aerobic respiration in larval stages to anaerobic respiration in adult stages to adapt to low oxygen microenvironments in host tissues [23, 24, 38]. Given that oxygen tension fluctuates within parasite lung cysts, adult P. westermani are thought to be facultative anaerobes with separate populations of mitochondria capable of either aerobic or anaerobic respiration [39, 40]. Pathway modules associated with aerobic respiration (e.g. M00087: beta-oxidation, M00009: citrate cycle, M00148: succinate dehydrogenase, etc.) were complete in both transcriptomes (Additional file 4: Table S3), and key enzymes involved in anaerobic dismutation (e.g. phosphoenolpyruvate carboxykinase) were also identified (Additional file 2: Table S1 and Additional file 3: Table S2). Modules related to fatty acid initiation (M00082, two of 13 KOs) and elongation (M00083, one of 14 KOs) are incomplete and poorly represented, so it is unlikely that these processes take place in adult Paragonimus (Additional file 4: Table S3), although fatty acid binding proteins were identified in both species (based on NR matches; comp22449_c0 and comp19053_c0 in P. westermani and comp74673_c0 in P. skrjabini). This is consistent with the hypothesis that trematodes (with the possible exception of C. sinensis [35]) are incapable of de novo fatty acid biosynthesis [23, 24, 41].

Host-parasite interaction

Secreted and excreted proteins are of particular interest in parasites like Paragonimus. They often play important roles in host parasite interaction [41, 42] and are useful targets for diagnostic assays [43–45]. While the N-terminal regions of proteins, which contain secretion signals, are often underrepresented in transcriptome assemblies, 622 P. westermani and 750 P. skrjabini unigenes were found to contain classical signal peptides and no transmembrane domains. This suggests that they may be secreted from cells. Several GO terms related to proteolysis and redox regulation were enriched in the putative secreted unigenes in both species (Additional file 5: Table S4). This is consistent with previous findings that highlighted the prevalence of proteases in trematode excretory-secretory products [46–49] and outlined their important roles in migration through host tissues, feeding and immune evasion [50–53].

Molecular mimicry is a well-known strategy for host manipulation and immune evasion [54]. Interestingly, 122 and 134 predicted proteins from P. westermani and P. skrjabini had far better blast matches to Homo sapiens (a potential host species) than to the free-living, freshwater planarian platyhelminth, Schmidtea mediterranea (Additional file 2: Table S1; Additional file 3: Table S2; see Methods for details). These putative “host mimic” proteins were enriched for kinase and GTPase activity in both species (Additional file 5: Table S4), which may indicate roles in signaling. Parasites like Plasmodium spp., Echinococcus multilocularis and Schistosoma mansoni are known to possess functional homologs of host hormone receptors [54–57]; thus there is a precedent for comingling of host and parasite signaling pathways.

Gene expression in Paragonimus westermani and P. skrjabini

Expression levels were estimated for each unigene in the two transcriptome assemblies (Additional file 2: Table S1; Additional file 3: Table S2). As expected, the top 5 % most highly expressed unigenes in both assemblies were enriched for GO terms related to basic cellular functions such as translation, ATP synthesis and redox regulation (Additional file 5: Table S4). Finding a direct one-to-one correlation between assembly unigenes can be challenging due to the incompleteness and fragmentation of de novo transcript assemblies; however, 12,432 P. westermani and P. skrjabini unigenes were linked through a bi-directional blast match of the longest transcript isoform from each. The expression of matched unigenes tended to be well correlated, but some differentially expressed unigenes were identified (Fig. 4, Table 4). The 303 unigenes that were upregulated in P. westermani were enriched with GO terms related to endopeptidase activity whereas the 249 unigenes upregulated in P. skrjabini were enriched with GO terms related to microtubule based movement (Additional file 5: Table S4). Disparities in gene complement and expression such as these could account for the striking biological differences between P. westermani and P. skrjabini.

Fig. 4
figure 4

Estimated expression of Paragonimus westermani and P. skrjabini assembly unigenes. Corresponding P. westermani and P. skrjabini unigenes identified by bi-directional blast search. The expression levels of each unigene (FPKM) were estimated, and the expression of matched unigenes were compared. a The average fold change plus or minus 1.96 times the standard deviation (corresponding to the top 5th percentile of differential regulation) was used as a cutoff to select unigenes differentially expressed between the two species. b Expression levels of corresponding unigenes were plotted, and differentially expressed unigenes are colored

Full size image
Table 4 Top 30 differentially expressed unigenes of P. westermani and P. skrjabini
Full size table

Diagnostic potential of deduced P. westermani and P. skrjabini proteins

In a previous study, proteins predicted from the de novo transcriptome of P. kellicotti were used as a comparative database in a mass spectrometry study aimed at identifying parasite proteins that could be used as serodiagnostic markers [16]. Paragonimus kellicotti proteins were immunoaffinity-purified from worm lysate with IgG from the serum of infected patients and proteins predicted from 321 transcripts (227 unigenes) were identified by mass spectrometry. Some 205 of the immunoreactive P. kellicotti proteins have blast matches to proteins deduced from the transcriptomes of both P. westermani and P. skrjabini (Additional file 2: Tables S1; Additional file 3: Table S2). Among these conserved proteins was a putative myoglobin isoform proposed as a diagnostic candidate due to its high detection levels in the MS study and its low sequence conservation with trematodes of other genera (Fig. 5). Further studies will be needed to thoroughly explore the utility of this protein as a pan-Paragonimus diagnostic marker.

Fig. 5
figure 5

Alignment of myoglobin orthologs from Paragonimus species and other trematodes. Although assembly fragmentation resulted in a truncated sequence from P. skrjabini, it had greater 90 % similarity with Paragonimus myoglobin (at the amino acid level), with much less similarity to myoglobins from other trematodes. Abbreviations: Pk, Pk34178_txpt1 [16]; Pw, comp20873_c0_seq2; Ps, comp80973_c0_seq3; Cs, C. sinensis gi:349998765; Ov, Opisthorchis viverrini gi: 663047528; Fh, F. hepatica gi:159461074; Sm, S. mansoni gi:256084837; Sj, S. japonicum gi:226487206

Full size image

Conclusions

This study provides the first insights into gene content and expression in P. westermani and P. skrjabini. Genetic conservation and diversification were assessed to characterize present and absent metabolic pathways. Like other FBTs [23, 24, 41], these species appear capable of both aerobic or anaerobic metabolism, but not de novo fatty acid biosynthesis. For the most part, conserved unigenes were expressed to similar degree in both species. Genes upregulated in P. westermani were enriched for GO terms related to proteolysis while genes upregulated in P. skrjabini were enriched for GO terms related to microtubule based movement. Expressed orthologs of P. kellicotti serodiagnostic antigens were identified in both species, and should be explored in pan-Paragonimus diagnostic assays. We expect that the assembled transcriptomes and the accompanying functional annotations will be a valuable resource for future research, including ongoing genome sequencing projects [33].