Deep, multi-stage transcriptome of the schistosomiasis vector Biomphalaria glabrata provides platform for understanding molluscan disease-related pathways

Summary statistics related to reads can be seen in Table 1. Read quality was found to be excellent, with lower quartile Phred scores above 32 through to the 101st base (Additional file 1: Figure S1). The mean GC content of reads, 39 %, is consistent with values observed in previous EST-based investigations, for example [14], where 38.54 % content was found. In the first 5 positions in our reads a number of sequences were found to be over-represented, likely to be a result of bias in adaptor hexamer binding as previously noted in other Illumina-based sequencing experiments [28] as no residual adaptor sequence was observed in our tests of the raw data.

Statistics on the final assembly can be seen in Table 2, with a graph of the distribution of sequence lengths seen in Fig. 3a. The compressed fasta file with our transcriptome assembly (≥500 bp) can be downloaded from Figshare with the doi 10.​6084/​m9.​figshare.​3117385, and the untrimmed assembly (≥100 bp) is available from 10.​6084/​m9.​figshare.​3117406 (to download directly, type https://​dx.​doi.​org/​ before the doi reference). In our 500 bp minimum length assembly we recover 295 Mbp of transcribed sequence, which provides a deeper resource to draw from than heretofore available in this species. With 90,315 contigs greater than 1 kb in length and an N50 of 3,221 bp, the dataset is well-assembled and will reliably span complete domains, allowing for easy identification of genes.

Assessment with the Core Eukaryotic Genes Mapping Approach (CEGMA) set of 248 ultra-conserved core eukaryotic genes (CEGs) found in nearly all eukaryotic organisms [42] identified 247 of these as present in our dataset. This represents 99.6 % coverage, a figure comparable or better than most complete genomic sequencing projects, and suggests that this transcriptomic resource provides an almost total recovery of the basic genetic repertoire of this species.

While sequencing efforts on the B. glabrata genome are still underway, transcriptomic resources are vital for gaining an understanding of the molecular processes behind aspects of this species’ disease resistance, development and physiology. To date, the most complete resource available, [14], is a fairly comprehensive transcriptomic resource, assembled from 30 adult snails, ranging from juvenile (4-7 mm shells) to older adult (12-16 mm shells) BgBRE individuals, from Illumina reads. However, this transcriptome does not sample as broadly across the course of development as the one presented here, and suffers according to many standard metrics of assessment. For example, the N50 of the Dheilly et al. resource, 1,067 bp according to our calculation, is threefold less than that of the assembly described here (3,221 bp). Given that 30,206 sequences were assigned a GO identity in the Dheilly et al. assembly, c.f. 74,492 in ours, the additional length of our assembly also is advantageous in allowing more firm homology assignment to be made.

To assess the degree of overlap between our sample and the Dheilly et al. resource we then used BLASTn (with the Megablast setting) to detect the number of sequences shared in common between the two assemblies. Of the 326,874 sequences in the ‘transcriptsBre1et2.fasta’ file, 241,019 (73.73 %) were found in our 500 bp minimum size assembly, and 268,903 (82.26 %) in our original, untrimmed dataset. Reciprocally, of the 133,084 contigs in our assembly, 119,264 (89.62 %) have at least a partial hit in the Dheilly et al. resource. We therefore recover the majority of the Dheilly et al. assembly in our data, with markedly more complete length as discussed earlier. We suspect that much of the portions of the transcriptomic resources that do not overlap are temporally or spatially restricted in expression, with those in our assembly representing embryonically transcribed genes, and those in Dheilly et al. more likely to be adult-specific.

BLAST2GO was used to functionally annotate this data for comparison with other species and previously completed B. glabrata GO distributions [13]. The results of this analysis can be seen in Fig. 3b and c, and full GO annotations are available in Additional file 2.

As a result of the excellent assembly statistics, a large number of identifiable sequences were obtained. Of the 133,084 contigs in our dataset, 80,952 (60.8 %) possess a homologue in the nr database (BLASTx, E value cut off 10^-3). Comparison of the BLASTx best hits results by species reveals that, despite the accessioning of oyster C. gigas and limpet L. gigantea genomic data onto the nr database, the closest hits gained using BLASTx are more heavily weighted towards deuterostome species. These species have previously been noted as having a slow rate of molecular evolution, and it may be this resultant similarity in sequence, compared with the fast-evolving Ecdysozoan species similarly well represented in the nr database, which causes this result.

Of the sequences with a BLAST-annotatable homologue in the nr database, 74,492 were mapped and of these 53,412 were assigned GO terms. The distribution of several GO assignment distributions in the three second-level functional categories (biological process, cellular component and molecular function), can be seen in Fig. 3c. In concert with our CEGMA results, we are therefore confident that this dataset contains the sizeable majority of transcribed RNA in this species, although it is likely a number of RNAs of low copy number and restricted temporal expression are not present.

The GO terms shown in Fig. 3c correspond to those shown in [13], a previously published annotated B. glabrata transcriptomic dataset. This allows comparison of the results of BLAST2GO analysis between our data, previously extant data for this species, and the distributions of the proteomes of the well-described model organisms D. melanogaster and M. musculus. The distributions seen in Fig. 3c suggest that our dataset is more representative of the GO term distributions seen in complete proteomes than those available previously in B. glabrata. While some differences are to be expected in the distribution of GO identities from species to species and GO distribution is an imperfect measure of the completeness of a dataset, our results, in red, mirror those of the fully sequenced species much more closely than those of the previously available B. glabrata datasets shown in blue (Fig. 3c).

Along with CEGMA and GO-based evidence, KEGG pathway mapping suggests we have near-100 % coverage of all major signaling and metabolic pathways, with well-conserved processes such as gluconeogenesis and the citrate cycle demonstrating complete coverage of all their constituent steps. The coverage of these, and other key pathways, can be seen in Table 3, and all annotations can be seen in Additional file 3. Generally, greater than 90 % coverage can be seen for all genes expected in protostome metazoans (NB, KEGG pathways shown also show genes restricted to other clades).

To illuminate the coverage provided by our dataset and its utility in reconstructing cellular processes involved in immune responses and developmental signalling, the Wnt, apoptosis and Notch signaling pathways are shown in Fig. 4. Apoptosis is a key part of molluscan disease response [50] and the Wnt and Notch pathways are involved in a wide range of patterning mechanisms in the process of growth and development.

While coverage of the KEGG apoptosis pathway map is the least well-recovered of the pathways shown here, at 77 % it is still remarkably complete, and contains many prospective targets for future research. The differences in orthology between spiralian genes and the more well-researched vertebrate paralogs (after the whole genome duplications observed in that lineage) will require careful unteasing. For example, protostomes exhibit far less CASP gene diversity, but sub- and neo-functionalisation in vertebrates may mean that single B. glabrata orthologues likely play the role of several paralogous sequences from the Vertebrata. The recovery of greater than three quarters of this pathway in B. glabrata here nevertheless represents a considerable advance in our knowledge of this pathway in the Mollusca in general, and in B. glabrata in particular.

Similarly, the Wnt and Notch signaling pathways are well-recovered in our dataset and are representative of similar signaling cascades not shown in figures here, but available in Additional file 3. Such recovery in key conserved signaling cascades suggests that our dataset could also be of utility for a range of further investigations in fields such as endocrinology, developmental and cell biology.

It should also be noted that some genes may be present in our dataset, but be considerably different in sequence to the query database used to identify genes, which currently contains a very limited number of spiralian gene sequences. This would lead to under-reporting of completeness, and lends further weight to hypothesized excellent coverage in our dataset in known gene pathways.

To assay the potential utility of this transcriptomic dataset as a resource for investigation of disease processes, we have targeted particular gene families for manual annotation. We are able to recover a range of disease-response associated genes in our transcriptome, including a number of genes from families not before identified in B. glabrata. These genes are listed in Table 4, and their complete sequences can be found in Additional file 4.

Molluscs generally utilize innate immune responses, such as cell-mediated reactions, to fight infection, as they appear to lack the adaptive immune system found in vertebrates. Numerous gene families were also examined and found to be present in this resource. Nuclear factor kappa-light-chain-enhancer of activated B cells (NF- κβ) genes are found in all metazoan lineages, and are noted as "rapid-acting" primary transcription factors, capable of quick response to harmful stimuli [19]. Two NF- κB have been previously published in B. glabrata: BgRel (ACZ25559.1) and BgRelish (ACZ25560.1) [59]. Both of these sequences are found in our dataset, with several forms of BgRelish gene identified (split between Comp98599, which possesses eight potential isoforms, and Comp97091, with three sequences recovered). Several variants of the BgRel gene were noted in our sample (Comp89844, four isoforms, and Comp91314, six isoforms). Two potential nuclear factor of activated T cells genes were also noted in our sample, Comp99436 (two isoforms) and Comp99884 (6 isoforms), which provide an additional target for functional work in this well-conserved gene family, and further suggests a role for Toll-like receptor (TLR) and immune deficiency (IMD)-like pathways in immune response.

CREB (cAMP response element-binding protein) and STAT (signal transducer and activator of transcription) genes play roles in proliferation and phagocytosis respectively [59]. Using known B. glabrata sequences as queries (tBLASTn, E < 10^-10), homologues of these genes are found in at least two and 19 contigs respectively in our dataset, and these sequences can be seen in Additional file 4, which will allow further work complementing previous investigations in this species [59]. The functional role for these novel sequences are as yet uncharacterized in B. glabrata or the Mollusca in general, but with the description of this complement, further investigation is possible.

Along with other components of innate immune responses, molluscs are known to utilise fibrinogen-encoding proteins (FREPs) as part of their immune defences [1, 58]. These contain immunoglobulin superfamily domains and can therefore be recognized even in de novo datasets [60]. Recent RNAseq-based efforts [14] have uncovered two related families of genes related to FREPS, known a C-type lectin-related proteins (CREPs) and Galectin-related protein (GREPs) which also are believed to play a part in mediating these responses. Collectively, these molecules are known as Variable Immunoglobulin and Lectin domain containing molecules (VIgL). The transcriptomic resource described here contains 81 contigs with high similarity (tBLASTn, E < 10^-9) to known B. glabrata VIgL proteins, along with many of slightly less strong homology. While complete mapping of the VIgL families is beyond the scope of this manuscript, this resource adds significantly to the number of these genes which can be investigated in detail for their responses to schistosome infection in the future.

For example, [56] used a microarray-based method to identify 98 differentially expressed sequences from haemocytes of schistosome-resistant and schistosome-susceptible B. glabrata after exposure to excretory-secretory products from S. mansoni. Of these 98 EST derived sequences, 61 were at that time unable to be assigned homology to known gene families due to their short length, and therefore were excluded from further analysis in that manuscript. However, using BLASTn (cutoff E < 10^-6, although this value generally = 0.0) we were able to recognize 20 more of these EST sequences within our transcriptomic dataset. With their greater average length (mean = 1359.9 bp) we were able to determine that one contig in our dataset (comp100220_c2_seq2) corresponded to five of the ESTs included in the list of 61 unknown sequences. This contig can, with the aid of the complete sequences provided by our dataset, be recognized as belonging to a single, long alpha-actinin A like-sequence, whose prevalence in earlier EST-based studies might indicate a cytoskeletal response to schistosome infection.

Furthermore, the additional sequence provided by our contigs allowed more of these unknown sequences to be more firmly identified. Of the 20 newly identifiable EST sequences (including all five matching comp100220_c2_seq2), only four were unable to be matched to known sequences in Genbank using BLASTx. In almost all cases, the best hits were also to B. glabrata sequence provisionally accessioned by other studies, with homologous genes in other species matching less well. Of these 16 ESTs identifiable courtesy of our resource, 5 are similar to alpha-actinin A as noted above. Others include gastric intrinsic factor, serine/arginine-rich splicing factor 7 and NRDE-2. These genes also may be useful targets in the treatment of schistosomiasis, and our dataset can in this way act as a bridge between previous functional investigations such as [56] and sequences of presently unknown function in the nr database.