Establishment and analysis of a reference transcriptome for Spodoptera frugiperda

Abstract

Background: Spodoptera frugiperda (Noctuidae) is a major agricultural pest throughout the American continent. The highly polyphagous larvae are frequently devastating crops of importance such as corn, sorghum, cotton and grass. In addition, the Sf9 cell line, widely used in biochemistry for in vitro protein production, is derived from S. frugiperda tissues. Many research groups are using S. frugiperda as a model organism to investigate questions such as plant adaptation, pest behavior or resistance to pesticides.

Results: In this study, we constructed a reference transcriptome assembly (Sf_TR2012b) of RNA sequences obtained from more than 35 S. frugiperda developmental time-points and tissue samples. We assessed the quality of this reference transcriptome by annotating a ubiquitous gene family--ribosomal proteins--as well as gene families that have a more constrained spatio-temporal expression and are involved in development, immunity and olfaction. We also provide a time-course of expression that we used to characterize the transcriptional regulation of the gene families studied.

Conclusion: We conclude that the Sf_TR2012b transcriptome is a valid reference transcriptome. While its reliability decreases for the detection and annotation of genes under strong transcriptional constraint we still recover a fair percentage of tissue-specific transcripts. That allowed us to explore the spatial and temporal expression of genes and to observe that some olfactory receptors are expressed in antennae and palps but also in other non related tissues such as fat bodies. Similarly, we observed an interesting interplay of gene families involved in immunity between fat bodies and antennae.

Figure 1

Content of the reference transcriptome. A. Barplot representing the percentages of multiple hit reads, unmapped reads and uniquely mapped reads, as provided by Bowtie, when aligning an RNAseq library against either the Sf_454_clusters assembly or the Sf_TR2012b assembly. The percentages obtained are the average of four independent experiments. B. Pie chart representing the number and percentage of contigs from Sf_TR2012b grouped by their best blastx hit against nr. Number of contigs and percentage of the total are represented. C. Synthetic table representing the number of genes found per family in the Sf_TR2012b assembly. The number of full and partial transcripts for Hox-domain proteins is irrelevant because the only conserved part is the homeodomain itself.

Figure 2

Schematic representation of S. frugiperda immune components found in Sf_TR2012b. The four main signaling pathways involved in insect immune response are detailed as well as the pro-PO cascade. The negative regulators are red circled and the components that were not found in Sf_TR2012b are indicated by red arrows.

Figure 3

Representation of transcripts and their expression. A. Screenshot of the GBrowse system integrated in the Lepidodb. A region centered around spod-11-tox [34] is represented on the 33K20_Sf BAC. The manual annotation of this already described gene can be compared with KAIKOGASS_mRNA predictions (top track) and Sf_TR2012b transcript (second to top track). The bottom two tracks represent coverage reads from 2 Illumina tissue RNAseq experiments, induced fat bodies and larval antennae and palps. B. Venn diagram showing the overlap between Kaikogass gene predictions on the BACS and Sf_TR2012b transcripts aligned on the same BACs. C. Correlogram of the rpm values for each of the developmental time points and tissue RNAseq experiments.

Figure 4

Clustering of expression. A. Heatmap representing the medoids of expression of the 20 clusters of genes for the 10 Illumina RNAseq experiments. B. Barplot representing the number of genes that are present in each cluster. C. Histogram representing the density of genes in each cluster for all (green) or RBP genes (red). Orange represents the intersection. It shows if RBP genes are over- or under-represented in each cluster compared to the total number of genes in each cluster. D. Same as in C. for Hox-domain genes.

Figure 5

qPCR validation of differential expression. A. Heatmap representing the expression in rpm of each candidate transcript tested by qPCR. Genes are ordered from top to bottom from a lesser ratio between eggs and L2e samples, as measured in qPCR to a higher ratio. The red box shows the 2 genes that we used to normalize the qPCR. B. Same heatmap as in A. but showing expression as z-scores scaled by row to highlight the differential expression between eggs and L2e. C. Barplot showing the ratio measured in qPCR, using elongation factor 3 as a negative control for normalization. D. Scatter plot showing the correlation between fold changes measured by RNAseq (y axis) and ratios measured by qPCR for the tested genes. The two measurements have a correlation coefficient of 0.74. A linear regression model has been applied and is also shown on the same graph.

Figure 6

Tissue specific expression of chemosensory genes and anti-microbial peptides. A. Heatmap showing the expression as row scaled z-scores of S. frugiperda odorant-binding proteins, chemosensory proteins, olfactory receptors and ionotropic receptors in the 10 Illumina RNAseq experiments. A higher expression of odorant-binding proteins in the antennae and palps can be observed. B. Same as in A. for the anti-microbial peptides identified in S. frugiperda transcriptome. A higher expression of AMP in induced fat bodies is observed as well as an overexpression in antennae and palps as well as in tracheae, two tissues in contact with the external environment.