To create and annotate a de novo transcriptome assembly for Antarctic krill a preliminary investigation focusing on the efficiency and quality of already existing strategies for de novo transcriptome assembly of non-model organisms was performed. In a second step, we focused on identifying and applying the best transcriptome assembly strategy to finally explore the gene expression levels across different developmental stages and krill responses to different environmental conditions. At first, separate transcriptome reconstructions using different assembly programs were carried out. A combination of two filtering steps was applied to these results to discard artifacts and improve the assembly quality. Reconstructed transcripts across all assemblers were joined, producing a set of non-redundant representative transcripts. We obtained these results by applying the EvidentialGene pipeline (version 4), which was specifically designed to combine different reconstructions and to eliminate redundant sequences. Finally, we applied another filter to identify redundant or mis-assembled sequences still appearing in the transcriptome.
Transcriptome quality
We checked the quality of our reconstructed transcriptome step by step, starting from the independent de novo assemblies, then evaluating the potential of merging all assemblies into a unique meta-assembly, and finally filtering the transcriptome for redundancy. All these results are summarized in Fig. 1, Tables 1 and 2. The result of our reconstruction strategy was evaluated using different measures: the N50 statistics highlighted an increase in transfrag lengths at each step. Recent benchmarks, such as18, have shown that, while reconstructing the transcriptome of a species, no single approach is uniformly superior: the quality of each result is influenced by a number of factors, both technical (k-mer size, strategy for duplicate resolution) and biological (genome size, presence of contaminants). In our study, we observed that, although a consistent number of sequences was removed through each step of the assembly, merging and filtering procedure, we didn’t encounter any decline in the quality described by the basic statistics of the reconstructed transcripts (Table 1).
Transcriptome quality assessment results. Results of the first assembly filtering in terms of total number of transcripts.
We then explored the completeness of the krill transcriptome according to conserved ortholog content using BUSCO (version 4.0.5) comparing our sequences to all the expected single-copy orthologs from the Arthropoda phylum. The results of the BUSCO analyses performed on each independent de novo assembly, on the EvidentialGene reconstruction and the final transcriptome are reported in Table 2. This analysis confirms that our strategy for controlling redundancy did not affect transcriptome completeness: indeed, the fraction of complete single-copy essential genes dropped by 1.8% only, while 123,376 redundant transfrags were discarded.
We finally compared our quality assessment results with those from previously released krill transcriptomes (Table 3). Our latest assembly significantly improves all the metrics we have discussed above. While this evidence suggests that our assembly is reasonably close to providing a complete representation of the krill transcriptome, it is more difficult to gauge the amount of redundancy it contains. Specifically, it remains difficult to distinguish between splice variants of a gene and possible paralogous copies. We believe that only the availability of a genome draft will make it possible to reliably discriminate between these two signals.
Functional classification
The assembled fragments were aligned against known protein and nucleotide databases to understand whether they could be linked to specific functions or processes described in other species. The functional annotation analyses showed that 63,903 contigs (42% of the total krill transcriptome) matched at least one protein from the NCBI NR (non-redundant) collection for a total of 98,316 unique proteins, while 62,518 transfrags found homology with a UniProtKB/TREMBL protein sequences (41% of the total), matching a total of 96,005 unique proteins. Furthermore, 22,024 krill transcripts (15% of the total) had significant matches with sequences in the NCBI NT nucleotide database. To classify transcripts by putative function, we performed a GO assignment. Specifically, 2833 GO terms (corresponding to 13,064 genes) were assigned: 1224 of those (corresponding to 11,575 genes) represented molecular functions; 1193 terms (corresponding to 6990 genes) were linked to biological processes; 416 terms (corresponding to 4301 genes) represented cellular components.
A case study on the discovery of opsin genes
To evaluate the gene discovery potential of the new assembly, we searched the transcriptome for novel members of the opsin family. Opsins are a group of light sensitive G protein-coupled receptors with seven transmembrane domains. Fourteen genes were annotated as putative opsins, and the conserved domains analysis revealed that all of them possess the distinctive 7 α-helix transmembrane domain structure. The eight previously cloned opsins20 were all represented in KrillDB2 (sequence identity > 90%; Table S1 Supplementary Material). The other six genes we identified can therefore be considered new putative opsins. Among those, we found four putative rhabdomeric opsins: EsRh7 and EsRh8, with 70% and 59% of amino acid identity to EsRh1a and EsRh4, respectively; EsRh9 and EsRh10 showing high sequence identity (87% and 74%, respectively) to EsRh5. Furthermore, we identified two putative ancestral opsins: a non-visual arthropsin (EsArthropsin), and an onychopsin (EsOnychopsin) with 70% and 49% of sequence identity with crustacean and onychophoran orthologous, respectively. Phylogenetic analysis (Fig. 2) suggested that EsRh7-10 are middle-wavelength-sensitive (MWS) rhabdomeric opsins, and further confirmed EsArthropsin and EsOnychopsin annotation.
Phylogenetic relationships of Euphausia superba opsins shown as circular cladogram. Colored dots indicate krill opsins: red, previously cloned opsins; green, novel identified opsins. The spectral sensitivities of rhabdomeric opsin clades were inferred from the curated invertebrate-only opsin dataset proposed by DeLeo & Bracken‐Grissom, 2020. Represented opsin classes: LWS, long-wavelenght-sensitive; LSM, long/middle-wavelenght-sensitive; MWS, middle-wavelenght-sensitive; SWS/UV, short/UV-wavelenght-sensitive; ONY, onychopsins; MEL, melanopsins; PER, peropsin; ART, arthropsin. Rectangular phylogram is reported in Fig. S1 (Supplementary Material).
Differential expression
The availability of a new assembly of the krill transcriptome, reconstructed by collecting the largest amount of experimental data available thus far, suggested the possibility of performing a more detailed investigation of differential expression patterns. Therefore, we decided to reanalyze the dataset from Höring et al.21 to assess the possibility of identifying differentially expressed genes that were not detected in the original study due to the use of an older reference transcriptome15.
Our design matrix for the model included all the independent factors (season, area and sex) and, in addition, the interaction between area and season, sex and area, sex and season.
In total 1741 genes were differentially expressed (DEG) among experimental conditions. They correspond to around 2% of the total reconstructed genes. In the previous work by Höring21, the same samples were quantified against 58,581 contigs15 producing 1654 DEGs. Table 4 summarizes the list of performed contrasts, each one with the number of differentially expressed up and down regulated genes.
1195 DEGs were identified in the comparison between summer and winter specimens: 1078 were up-regulated and 117 down-regulated. In addition, 396 of such DEGs had some form of functional annotation. In general, these results are in accordance with the discussion by Höring21, which found that seasonal differences are predominant compared to regional ones. A summary of the DEGs is listed in Table 5. Complete tables of differentially expressed genes are downloadable on KrillDB2 (Fig. 3c; https://krilldb2.bio.unipd.it/, Section “Differentially Expressed Genes (DEGs)”).
Blast search section. The new search box for sequence searches (a) with an example of a BLAST search (highlighted in yellow) and the corresponding results (b). By clicking on each target identifier, the user will be redirected to that specific transcript page, where new sections have been added, as shown in Fig. 6.
Summer versus winter
We selected a series of genes among seasonal DEGs according to what has been already described in the literature. Höring et al.21 previously identified and described 35 relevant DEGs involved in seasonal physiology and behavior: we recovered the same gene signature in our analysis by comparing summer to winter samples. The majority of these DEGs appear to be involved in the development of cuticles (chitin synthase, carbohydrate sulfotransferase 11), lipid metabolism (fatty acid synthase 2, enoyl-CoA ligase), reproduction (vitellogenin, hematopoietic prostaglandin D synthase), metabolism of different hormones (type 1 iodothyronine deiodinase) and in the circadian clock (cryptochrome). Our results also include DEGs that were involved in the moult cycle of krill in other studies16. Specifically, we identified a larger group of genes involved in the different stages of the cuticle developmental process (peritrophin-A domain, calcified cuticle protein, glycosyltransferase 8-domain containing protein 1, collagen alpha 1, glutamine-fructose 6 phosphate), including proteins such as cuticle protein-3,6,19.8, early cuticle protein, pupal cuticle protein, endocuticle structural glycoprotein, chitinase-3 and chitinase-4, the latter representing a group of chitinase which have been shown to be expressed predominantly in gut tissue during larval and/or adult stages in other arthropods and are proposed to be involved in the digestion of chitin-containing substrates22. Finally, in addition to trypsin and crustin 4 (immune-related gene, essential in early pre-moult stage when krill still have a soft cuticle to protect them from pathogen attack, as seen by Seear et al.16), we also identified crustin-1,2,3,5 and 7. All the reported genes were up-regulated in summer, the period in which growth takes place and krill moult regularly.
Cuticle development genes were also identified as differentially expressed in the analysis of the interaction of multiple factors, between male samples coming from South Georgia and female specimens coming from the area of Bransfield Strait-South Orkney (considered as a unique area since they are placed at similar latitudes). Strikingly, we also identified a pro-resilin gene, whose role in many insects consists in providing efficient energy storage, being up-regulated in South Georgia male specimens.
Interaction effects
A number of relevant DEGs were found among specific regional and seasonal factors interactions. For instance, by comparing krill samples coming from South Georgia in summer and individuals sampled in Bransfield Strait-South Orkney in winter, we found genes up-regulated in summer in South Georgia related to reproductive activities, such as doublesex and mab-3 related transcription factor. The latter is a transcription factor crucial for sex determination and sexual differentiation, which was already described in other arthropods23. Since no differentially expressed gene related to reproduction was found by Höring et al.21 in the same comparisons, this suggests that the new krill transcriptome improves the power to identify new expression patterns and characterize the krill samples.
Finally, the comparison between male individuals from the Lazarev Sea and female specimens from the Bransfield Strait-South Orkney showed additional DEGs involved in reproduction, such as ovochymase 2, usually highly expressed in female adults or eggs, serine protease and a trypsin-like gene. In particular, trypsin-like genes are usually thought to be digestive serine proteases, but previous works suggested that they can play other roles24; many trypsins show female or male-specific expression patterns and have been found exclusively expressed in males, as in our analysis, suggesting that they play a role in the reproductive processes.
The simultaneous presence of differentially expressed genes involved in different steps of the krill moulting cycle, in the reproductive process and in sexual maturation that appear to be differentially expressed in the same comparisons is in accordance with what was already observed in krill25 and other krill species26. In particular, there is evidence of a strong relation between the krill moulting process and its growth and sexual maturation during the year, which supports and confirms the reliability of our results in terms of genes involved in such krill life cycle steps.
Identification of microRNA Precursors
Although microRNAs play a key role in the regulation of gene expression and in many important biological processes, such as development or cell differentiation, there is still no information about microRNAs in krill species.
Here we performed an investigation to test whether the new transcriptome could also include sequences with a significant homology to known mature microRNAs.
In total we identified 261 krill transcripts whose sequences are highly similar to 644 known microRNAs from other species. 306 sequences were linked to at least one GO term, matching 54 krill transcripts (Table S2, Supplementary Material). Among them, we identified 5 putative microRNAs involved with changes in cellular metabolism (age-dependent general metabolic decline—GO:0001321, GO:0001323), as well as changes in the state or activity of cells (age-dependent response to oxidative stress—GO:0001306, GO:0001322, GO:0001324), 35 microRNAs involved in interleukin activity and production. We found 26 putative microRNAs likely involved in ecdysteroidogenesis (specifically GO:0042768), a process resulting in the production of ecdysteroids, moulting and sex hormones found in many arthropods. In addition, we found a microRNA involved in fused antrum stage (GO:0048165) which appears to be related in other species to oogenesis. We also identified 27 microRNAs related to rhombomere morphogenesis, formation and development (GO:0021661, GO:0021663, GO:0021570). These functions have been linked to the development of portions of the central nervous system in vertebrates, which share the same structure of those found in arthropod brains. Lastly, 26 krill sequences showed high similarity with 2 mature microRNA related to the formation of tectum (GO:0043676), which represents in arthropods and, specifically, crustaceans, the part of the brain acting as visual center.
KrillDB2 web Interface
The KrillDB website has been redesigned to include the new version of the transcriptome assembly. Figures 3, 4, 5 and 6 collect images taken from the new main sections of the database. The integrated full-text search engine allows the user to search for a transcript ID, gene ID, GO term, a microRNA ID or any other free-form query. Results of full-text searches are now organized into several separate tables, each representing a different data source or biological aspect (Fig. 5). Results of GO term searches are summarized in a table reporting the related genes with corresponding domain or microRNA match and associated description. Both gene and transcript-centric pages have been extended with two new sections: “Orthology” and “Expression” (Fig. 6). The Orthology section summarizes the list of orthologous sequences coming from the OMA analysis, each one with the species it belongs to and the identity score.
Differential Expression section. The new section collecting all differentially expressed genes tables (a) with an example of the corresponding result for a selected contrast (b).
New search engine of KrillDB2. Example of the results of a full-text search on KrillDB2.
Additional sections in gene and transcript pages. The new sections in the gene-centric page show a table listing the orthologous sequences with their belonging species and the identity score (a), a visualization of the gene structure as estimated by Lace software (d) and a boxplot coming from Expression Atlas analyses (c). Both Orthology and Expression sections are integrated also in the transcript-centric page. When a transcript is annotated as a putative microRNA, a “Predicted Hairpin” section displays a visualization of the hairpin predicted secondary structure and tables showing the alignment length, the HHMMiR score and the list of mature microRNAs matching (b).
The “Expression” section shows a barplot representing abundances estimates obtained from Salmon. An additional section, called “Gene Structure” (Fig. 6), was added to the gene page on the basis of the results coming from the SuperTranscript analysis. Specifically, we modified the STViewer.py Python script (from Lace), optimizing and adapting it to our own data and database structure, in order to produce a visualization of each gene with its transcripts. Since Lace relies on the construction of a single directed splice graph and it is not able to compute it for complex clusters with more than 30 splicing variants, this section is available for a selection of genes only.
The new KrillDB2 release includes completely updated transcript and gene identifiers. However, the user searching for a retired ID is automatically redirected to the page describing the newest definition of the appropriate transcript or gene.
The KrillDB2 homepage now includes two additional sections: one is represented by the possibility to perform a BLAST search (Fig. 3). Any nucleotide or protein sequence (query) can be aligned against krill sequences stored in the database. Results are summarized in a table containing information about the krill transcripts (target) that matched with the user’s query, and the e-value corresponding to the alignment. The other new section, called “Differentially Expressed Genes”, allows the user to browse all the tables listing the genes that were found to be differentially expressed among the conditions we have described above (Fig. 4). A drop-down menu gives access to the different comparisons; DEG tables list for each gene its log fold-change, p- and FDR values as estimated by edgeR. Moreover, each gene is linked to a functional description (if available) inferred from sequence homology searches.
Information about krill transcripts showing homology with an annotated microRNA is available in the “Predicted Hairpin” (Fig. 6). It contains a summary table with details about the hairpin length and the similarity score (as estimated by HHMMiR), followed by full listing of all the corresponding mature microRNAs (including links to their miRBase page). In addition, an image displaying the predicted secondary structure of the hairpin is included (computed by the “fornac” visualization software from the ViennaRNA suite).
Source: Ecology - nature.com
