Next-generation sequencing and hybrid contig assembly

Using Roche 454 FLX technology, two sets of libraries (MGE011: 122,471 reads; cDN18: 327,209 reads) consisting of a total of 449,680 reads were sequenced using normalized cDNA libraries constructed using either a mixture of adult tissues or containing gonadal tissue and entire larvae. The same libraries were used to obtain respectively 2,434 and 2,077 ESTs with traditional Sanger sequencing analysis. Using all data available, amounting to a total of 454,191 reads plus ESTs, a assembly was performed and grouped them into 41,119 contigs.

The average read size from 454 sequencing was 257 bp (Figure 1 shows the distribution of sequence lengths) and quality level of the reads was assured by a distribution of the sequences with 96% of reads with Phred sequence quality >20 (Figure 2, left panel) while ESTs of R. decussatus have a mean length of 604 bp. The GC content of the reads was in average 33% ± 6,6% similar with the EST sequences of R. decussatus deposited in Genbank (34.78 ± 6.26%), with a maximum of 74% (Figure 2, right panel). All 454 reads have been deposited in GenBank (SRA) with the accession number [SRA058431].

Transcriptome annotation and microarray quality assessment

To determine the putative identities of assembled contigs, Blastx and Blastn similarity searches on several protein and nucleotide sequence databases were performed. Of 41,119 unique sequences, 8,560 (21%) showed at least one significant match (e<10^-3) in the NCBI non-redundant protein database (Additional file 1). In addition to the annotation with Blast2GO, Blast searches against UniProtKB/Swiss-Prot database, UniProtKB/TrEMBL database and 5 different species-specific data bases (Additional file 1) were implemented in order to further increase the number of putatively annotated R. decussatus contigs (see Methods for details). This approach provided a significant match for additional 3,919 transcripts, which previously showed no correspondence with the NCBI non-redundant protein database, bringing the final number of clam entries associated with a known protein or transcript to 12,479 (30.3%). The percentage of annotated expressed sequences is very similar to that obtained for R. philippinarum (30%) by Milan and co-workers [24]. The highest number of significant similarity scores (9,364 hits, 22.8%) was obtained with Crassostrea gigas protein database, second best-matching species Lottia gigantea (8,828 hits, 21.4%) and third Danio rerio (7,170 hits, 17.4%) in accordance to what was previously observed for R. philippinarum annotation (24,1% and 18.5% annotated contigs with L. gigantea and D. rerio respectively).

Probe design took into consideration all annotated entries (12,479). Non annotated transcripts with sequence lengths ≥400 bp and average Phred sequence quality > 30 were also considered. A total of 21,900 target sequences were obtained and for each of them, two probes with opposite orientations (sense and antisense) were designed. A total of 43,758 out of 43,800 (99.9%) probes were successfully obtained, representing 21,887 R. decussatus transcripts. The percentage of annotated transcripts represented in the microarray was 57%. Annotated genes were categorized according to Gene Ontology (GO) Functions in the three root categories and also in terms of main families of genes using a previously defined GO classification (see Additional file 2). Probe sequences and other details on the microarray platform can be found in the GEO database under accession number GSE36276.

Validation of microarray data by quantitative real-time PCR analysis of gene expression

Comparative analysis of microarray and qPCR expression data is presented in Figure 3. Pearson correlation indicated coefficients of 0.984 for 12 compared genes and a p <0.005 when comparing data for microarray probe and correspondent qPCR, corroborating the good reliability of the microarray platform. Fold change is always higher by qPCR than measured by microarray with one exception, a common situation with Agilent microarrays [27].

Microsatellite content (SSR-EST)

The genome of Bivalves is known to harbor a large number of microsatellites [28], that can be useful for as markers for different kinds of studies such as population genetic structure, demography, selective breeding and quantitative trait loci studies in clam [29]. Nevertheless microsatellite marker development can be difficult to develop in mollusk for many reasons, some of them still unknown [30]. EST derived microsatellites have some advantages such as being more conserved across species and be more adequate for selective pressure studies for example [31]. Previous studies have characterized some EST-SSR in clams using 454 [26, 32] but none used R. decussatus as a model.

Rdecusdb, a Ruditapes decussatusdatabase

Rdecusdb (http://morse-ccmar.ualg.pt/edge) is centered on contigs sequence and annotation. All contig sequences as well as different layers of results for data analysis will be available through Tripal [35]. Tripal is an open source and freely available collection of Drupal modules for management and visualization of data stored within a GMOD Chado database. Analysis results are indexed by Drupal’s full text searching mechanism, allowing the users to find data of interest. For each contig, a gene-like entry shows different data and bioinformatic analysis results, being identified with a description together with a sequence in fasta format along with blast hits and the reads that assembled that same contig. In addition, for each contig and whenever predicted, Gene Ontology is given for Biological Process (BP), Molecular Function (MF), and Cellular Component (CC). IT includes several analysis such as: the Analysis BLAST homology module, the Analysis InterPro module, the Analysis KEGG module; and the Analysis GO module for displaying trees and charts for GO mappings. Recently the sequences were update due to a new assembly. The assembly used in this article and the recent one can be download from the website.

Comparison of gene expression in infected versus non infected clams

Clam gills were used because they are the main connection with the outside environment, together with the siphons which are one of the most affected tissues upon Perkinsus parasitism [36]. They participate as defense barriers sharing functions in the respiratory process and being also involved in the elimination of ROS molecules by endogenous antioxidant genes [37].

Data captured from transcripts fluorescence hybridization derived from four non infected and four infected clam gills was normalized and used to identify transcripts differentially expressed between the two conditions. Principal component analysis of conditions proved the good clusterization of samples (4 biological replicates) and consistence of the results between replicates, allowing the differences in steady-state mRNA levels between infected and non infected clams to be reliably measured.

All microarray data was deposited in the GEO database [38] under accession numbers GSE36276.

To perform microarray analysis a two unpaired class Significance Analysis of Microarray (SAM) test was carried out on normalized data. By imposing a False Discovery Rate (FDR) of 5% and Fold Change (FC) >1.5, a list of 949 probes, was obtained (see Additional file 3). From these, a total of 227 transcripts were up-regulated in infected clams versus non infected clams with a FC ranging from 1.5 to 102 while a total of 722 transcripts were down-regulated with a FC ranging from 1.6 to 76.

Genes were identified and categorized in terms of percentage using GO categories (cellular component, biological process and molecular functions) and specifically according to possible role in the immune response. For up regulated genes (Figure 4), more than half were found to be involved in general metabolism (53%) and in protein, lipid and carbohydrate metabolism (13%, 6% and 4% respectively) and in specific processes like stress response (12%) and response to biotic stimulus (4%). Expression of several genes associated with mitochondria represented two per cent of genes up regulated. Among the genes found to be down regulated, 46% were associated with general metabolism while only 1% was associated with lipid metabolism. For both categories, percentages were lower than those found for up regulated genes. Other gene clusters were also less represented when compared, such as those related with biotic stimulus (1.2%), stress response (7%) or even not present such as mitochondria related.

In contrast, the percentage of down regulated genes associated with apoptosis (1%), defense (1%), immune response (1%) and response to external/internal stimulus (2%) increased when compared with up regulated genes (Figure 5). Altogether, data suggests that metabolic and stress related genes are the most affected by Perkinsus parasitism, reflecting changes in growth and clam survival.

From all genes found to be differentially expressed upon Perkinsus infection (Additional file 3), only 67 of a total of 227 up regulated genes and 259 from a total 722 down-regulated genes were annotated successfully based on NCBI (National Centre for Biotechnology Information) amino acidic non redundant (nr) database. Some annotation was trivial and a second search using ncbi nr together with nt database was conducted, obtaining a more accurate annotation, returning 44 genes for up-regulated (Additional file 4), and 231 for down-regulated (Additional file 5). From these two lists we can identify lectins and a number of genes associated with immune/stress response. Up regulated genes include matrilin, a gene already associated with zebra mussel hemocytes host defense [39], methionine-r-sulfoxide reductase, a gene linked with antoxidant stress [40], exosome component 5, a set of genes capable of an immunomodulatory activity [41, 42], different kinds of proteases (serine proteases), some already described as defenses against Perkinsus in different bivalves [43–49], acid phosphatases-like genes and dimethylarginine dimethylaminohydrolase, a gene associated with immune response in amphioxus [50].

Among the down regulated genes (see Additional file 5), we can point out several related to calcium binding such as calmodulin. This finding is in agreement with previous data. Indeed, one of the majors players during Perkinsus infection is hypoxia [51] and calmodulin is known to be down regulated during hypoxia events in mussels [52]. Calmodulin can be associated with almost all cellular processes, including apoptosis, metabolism, inflammation and the immune response. Some immune responses such NF-κB signaling pathway in pearl oyster are regulated by calmodulin binding proteins such as calcineurin [53] and calmodulin was shown to be an important molecular determinant response to Perkinsus infection [54, 55].

The presence of lectins among the down-regulated genes constitutes an indication that Perkinsus parasitism can also negatively affect expression of some lectins as already shown for mussels [56] where multiples genes involved in immune defense are down regulated upon exposure to an infectious agent. Glutathione s-tranferases (GST) are also less expressed in this situation, a result contrary to our expectations since due to their role in cell detoxification and oxidative stress response, necessary for protecting the clam from the oxidative burst, we expected these levels to be up-regulated. However, and in agreement with our findings, some authors already demonstrated that in other mollusks, GST is up-regulated during the initial steps of infection but decreases when infection is established [57, 58]. Other genes related with oxidative stress, such glutaredoxin a, aldo-keto reductase, hmgb-like protein, hephaestin, glucose dehydrogenase, peptide o-xylosyltransferase-like or agglutination (hemagglutinin amebocyte aggregation [59]) and anti-apoptosis related genes (e.g. achain structure of the ciap2 ring domain) were also down regulated, supporting the theory that after the initial infection period there is a relaxation of immune defenses.

In order to obtain a more systematic functional interpretation of the set of differentially expressed genes, enrichment analyses using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) was performed. Indeed, among the up-regulated cluster of ESTs, we found a high number of transcripts coding for immune response related genes as expected. The immune system of clams and bivalves in general is deprived of an adaptive system and fight pathogen aggression through an innate immune response [60] exerted by humoral factors and cell-mediated mechanisms (Figure 6). Humoral factors include lectins (agglutinins, opsonins), lysosomal enzymes (phosphatase acid, lysozyme and various hydrolytic enzymes), antimicrobial peptides and protease inhibitors, among others [61]. The constitutive or induced expression of such genes can potentially be directly linked to an effort to arrest Perkinsus infection as observed in other systems. It is interesting to observe that oxidative processes, hydrolase activity and nucleic acid binding are the main processes represented and most were already pointed out as having influence during resistance to microorganisms (Table 2). At the cellular level we could also highlight a number of genes linked to non-membrane bounded organelles and to intracellular non-membrane bounded organelles, suggesting that activity from these organelles (ribosome, cytoskeleton related) is necessary for the entrapment of microorganisms and for protein synthesis.

Importance of lectins on Host Parasite interaction

Due to the important role of lectins upon Perkinsus infection, this family of genes was chosen to be the subject of a particularly detailed characterization. Three different families of lectins were identified: C-type lectins, C1q-lectins and galectins. The members of these families were identified using conserved residues as patterns for search. A search into Rdecusdb revealed over than 90 sequences using lectin as keyword, but some genes were just classified as having a lectin domain. For classification purposes all the transcripts were organized in families and superfamilies and respective fold change annotated. Also the domain architecture and presence of signal peptide was deduced (see Additional file 6: figure S6). Nevertheless 9 of those genes (Figure 7) were differentially expressed when comparing infected with non infected clams, one of them being the identified gene with the highest fold change in the array (over 50 times over expressed).

C-type lectins presented some conserved domains such WxD (position 139 and 141) [62] and CE (position 154) as shown in Figure 8. Cysteines are maintained due to their capability of forming disulphyde bonds [63], while WxD region is associated with calcium binding. Most of the lectins known to be linked to Perkinsus response are C-type lectins [10, 64]. They have diverse carbohydrate specificities and present multiple structural domains. They can be classified in different groups including collectins, proteoglycan core proteins, selectins, endocytic receptors, and the mannose-macrophage receptor, some of them directly or indirectly involved in immune function [65].

Galectins, which constitute one family of lectins, are characterized by a conserved sequence motif in their carbohydrate recognition domain (CRD) and a specific affinity for b-galactosides. When compared with other bivalves galectins, Ruditapes decussatus galectins present some conserved signature residues F(D/N)XR(F/L/I), (N/K)X(V/I/L)XXN and WGXERXR [66]. Interestingly, despite the large evolutionary distance between invertebrate and vertebrates galectins, they still share more than 30% of homology. Host galectins are also known to be related to presence of parasite and in oysters it was previously shown that some galectins are induced following Perkinsus marinus infection [12].

C1q domain containing proteins are essential in the innate immune system of invertebrates, and can be the link between innate immune system and adaptive immunity as seen in lamprey, where mammalian homologous C1q was shown to be act as a lectin in lamprey [67]. The C1q has a globular domain [68] and is one of the most represented lectins. C1q can be involved in a variety of immune related processes such as activation of complement system or pathogen recognition and even mediating cell migration [69]. The R. decussatus Sialic binding lectins (C1q) namely rud_dec_c29141 were shown to be highly expressed in the presence of Perkinsus olseni. To improve c1q pattern recognition all sequences from bivalve C1q lectins deposited in Genbank were collected (August 2012) and conserved residues were analyzed, allowing us to identify a bivalve C1q signature (Figure 9).

Sampling, cDNA library construction and sequencing

Samples of R. decussatus were collected in Faro, in the Ria Formosa lagoon system which spreads along the mid region of the southern Portugal coast. Total RNA was extracted using the acid guanidinium thiocyanate-phenol-chloroform method [74]. Two libraries were constructed, one using a mixture of all adult tissues from 20 individuals and a second using gonadal tissues from both sex with a ratio of 1 male to 4 females (gonadal phase IV) and juveniles clams with sizes ranging from 2 to 4mm total length. The cDNA libraries were constructed using the SMART kit from BD Biosciences Clontech and equal amounts of RNA and then normalized using the duplex-specific nuclease (DSN) method [75].

Sequencing was performed at the Max Planck Institute using 454 GS FLX instrument with Titanium series chemistry following manufacturer protocol. Pyroluminescence intensity was converted to sequence data using Newbler suite. 454 reads were post processing using sff_extract (0.2.8) and trimmed using clean_reads (0.2). Final reads quality was assessed using prinseq-lite (0.14.4).

Transcriptome assembly

A hybrid assembly using ESTs collected from Genbank and all 454 reads was performed to improve the assembly. For the latter the quality score files were taken into consideration. The purpose of a hybrid assembly is to explore the advantages of the two technologies, the numerous reads of 454 and the quality and length size of Sanger reads. MIRA3 performed the assemblies in two runs [76], where all contigs obtained with the first run of hybrid assembly were used for a second run to eliminate contig redundancy.

Transcripts annotation

Little information, specifically on gene annotation, is available in public databases for mollusks species, with the exception of recent deposition of sequences from Pacific oyster [23, 77], blue, Mediterranean [56] and deep sea vents mussel [78] and manila clam [24, 79]. Although those species are not annotated in a satisfactory way, they can still provide some extra information. Blast searches were conducted against NCBI (National Centre for Biotechnology Information) amino acidic non redundant (nr) database (release of March 2012), using Blastx option. Alignments with an E-value of at most 1.0 e^-10 were considered significant, and up to 10 hits per contig were taken into account.

Unfortunately, like any non-model organism, the annotation of the clam transcriptome can be a challenge and the annotation project was conducted using two other different strategies by i) blasting against ensemble protein databases of different species including Danio rerio, Gasterosteus aculeatus, Oryzias latipes, Takifugu rubripes, Tetraodon nigroviridis, Homo sapiens, Drosophila melanogaster using a cutoff value of <1.0 e^-5], and ii) blastn search (cut off e-value of <1.0 e^-5) against Lottia gigantea v1.0 database [80], Crassostrea gigas transcripts databases [23, 81] and Argopecten irradians EST database [82].

For de novo annotation of R. decussatus contigs, we used Blast2go tool, which encompasses all the tools for functional annotation of (novel) sequences and the analysis of annotation data [83, 84]. The Gene Ontology (GO) terms associations for BP, MF and CC were performed using Blastx algorithm against the NCBI amino acid nr database implemented in Blast2GO software. To categorize the GO terms into different GO categories, a web-based tool, CateGOrizer [85], was employed.

EST-SSR search

Misa software was used to screen simple SSRs and msatcommander for complex forms (combinations of different SSRs of coexistence of two or more SSRs). In either cases search was performed to obtain SSRs longer than 20 bp and at with the following tuning: dinucleotide repeat ≥ 20 bases; trinucleotide repeat ≥ 21 bases; tetranucleotide repeat ≥ 20 bases; pentanucleotide repeat ≥ 20 base; hexanucleotide repeat (HNP) (and more) ≥ 24 bases.

DNA microarray design

Agilent technology of oligo-DNA microarray was chosen to design a specific microarray based on sequenced transcriptome (Figure 10), containing two probes with both orientations considering all annotated transcripts and unknown transcripts with Phred quality>30 and length>400 bp. This 60mer oligo-probes design, in a 4 × 44K format, was assisted using the Agilent eArray interface [86].

Biological handling, RNA extraction, labeling and hybridization

Fifty grooved carpet shell clams with a size of 25-28 mm were collected from the wild in the Ria Formosa, Portugal and rested in proper aquariums for 4 days before sacrifice in order to reduce stress. Gills were dissected and immediately homogenize in Tri-reagent (Ambion, Austin. USA) and simultaneous a small portion of the gills and the rest of the tissues was incubated in Ray's fluid thioglycollate medium assay, following previously established protocol [4], to determine the level of Perkinsus.

Two groups were selected, one of four clams not infected (No hypnospores present) and another with four clams heavy infected with hypnospores presented in all tissues (Mackin scale 5). RNA was extracted individually from these two groups using Tri-reagent (Ambion, Austin. USA), following manufacturer’s instructions and later purified and treated with Dnase I using the RNeasy Mini Kit (Qiagen, Hilden, Germany), following the manufacturer’s instructions for RNA cleanup. RNA concentration was determined using a NanoDrop® ND-1000 spectrophotometer, (NanoDrop Technologies, Wilmington, USA) and integrity and quality were finally evaluated on an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA).

Labeling was done using 200 ng of total RNA linearly amplified and labeled with Cy3-dCTP (Agilent One-Color Microarray-Based Gene Expression Analysis). For control, a mixture of 10 different viral poly-adenylated RNAs (Agilent Spike-In Mix) was added to each RNA sample before amplification and labeling. Labeled cRNA was purified with Qiagen RNeasy Mini Kit, and sample concentration and specific activity (pmol Cy3/μg cRNA) were determined using a NanoDrop spectrophotometer. A total of 1,650 ng of labeled cRNA was prepared for fragmentation adding 11 μl 10X Blocking Agent and 2.2 μl of 25X Fragmentation Buffer, heated at 60°C for 30 min, and finally diluted by adding 55 μl of 2X GE Hybridization buffer. A volume of 100 μl of hybridization solution was then dispensed in the gasket slide and assembled to the microarray slide (each slide containing four arrays). Slides were incubated in the oven overnight at 65°C and then washed according to manufacturer’s protocol.

Microarray scanning and data processing

Scanning was performed twice at two different sensitivity levels (XDR Hi 100% and XDR Lo 10%) at 5 μm resolution using an Agilent G2565BA DNA microarray scanner. The two images were analyzed together and data were extracted and background subtracted using Agilent Feature Extraction (FE) Software version 9.5.1. After quality measures, all control features (positive, negative, etc.), except for Spike-in (Spike-in Viral RNAs), were excluded from subsequent analyses. Normalization procedures were performed using R statistical software using Spike-in control intensities to normalize each dataset. Significance Analysis of Microarray (SAM) [87] was used to identify differentially expressed genes between healthy clams and those infected with Perkinsus.

Functional enrichment of differentially expressed genes

Gene functional annotation based on gene enrichment was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.7 [88, 89]. DAVID is capable of recognizing functional annotation data from limited species, mainly human, mouse and zebrafish. So, in order to use clam data, it was necessary to convert clams genes into the equivalent orthologs of zebrafish Gene IDs or entrez entries, by blasting clam nucleotide sequences against zebrafish protein counterparts. This was done by downloading blasted annotated protein sequences and performing in-house blast routines. Then gene ontology search was performed in DAVID using two lists, one with all identified genes as background and another with the differentially expressed up and down regulated genes using the same predefined settings.

Multiple sequence alignments

Available c-type lectins and c1q sequences were retrieved from bivalve species present in Genbank, including Mytilus sp., Haliotis sp., Chlamys farreri, Mercenaria, and Ruditapes philippinarum and aminoacid protein sequences aligned using T-Coffee server [90], applying default settings. Alignments were subject to a posterior manual adjustment. C1q signature was determined using PRATT [91] server and sequence logos were then created from multiple alignments using WebLogo [92].

Architecture domain analysis

Domain analyses of lectins present in R. decussatus was performed using Superfamily at http://supfam.cs.bris.ac.uk/SUPERFAMILY[93] using nucleotide sequences of putative lectins discovered in our database. Lectins were classified according to domain architecture. Signal peptide was predicted using SignalIp 4.0 server [94].

Quantitative real-time PCR analysis