The different ESTs were classified into categories such as immune-related, housekeeping, hypothetical insect proteins, hypothetical non-insect proteins based on the homology in NCBI protein BLAST (Table 1). Of the 1412 ESTs, 432 (31%) showed similarity to known or putative insect immune proteins; thus, even though the cDNA library was not normalized, a significant proportion (one-third) of the transcriptome was represented by putative immune-related genes. A total of 569 ESTs (39%) were homologous to housekeeping genes and proteins involved in functions other than immunity. Functional class distribution of the ESTs is depicted in Table 1. Based on the sequence similarity, 679 of the 1412 ESTs were classified as insect-specific (224 of the 719 clusters) and 409 ESTs (204 clusters) were common to both insects and mammals (Figure 2a and 2b). The number of ESTs showing similarity to only the mammalian proteins was 14 and those showing similarity neither to insects nor to mammals were as few as 13. Gene Ontology analysis of the EST sequences was also carried out and the details are provided as supplementary data (see Additional data file 1).

The sequences, which could not be assigned any function based on homology search in NCBI, were searched for conserved domains in ProDom database. Of the 260 clusters that had no matches with known proteins in NCBI BLAST, we could assign protein domain families to 196 clusters based on the ProDom search (see Additional data file 2). The remaining sequences did not show any hits in ProDom and should be further analysed by other specialized computational tools.

We further screened the unannotated proteins (no hits or hits with hypothetical proteins in NCBI database) for the presence of signal peptide and absence of transmembrane domains. This characteristic of many of the immune proteins has been utilized to screen probable immune-related genes in large-scale transcriptome studies 17, although the reliability of this criterion to identify the immune related genes remains to be experimentally tested. In the present study, 25 out of 260 genes tested fulfilled these criteria and could be considered as potential immune-responsive genes. However, since all the 260 clusters checked are not all full-length sequences, it is possible that we may have missed some others amongst them that represent gene products harbouring signal peptide without transmembrane domain, and hence the actual number may be higher.

A total of 80 clusters were assigned a putative immune function based on their similarity with previously characterized immune response genes and their distribution pattern is shown in Figure 3. These putative immune proteins were categorized into different functional groups – i. anti-microbial proteins, ii. pattern recognition receptors (PRRs), iii. proteases and protease inhibitors and iv. putative defence proteins with unknown function. A subset of these putative immune genes (thirty-eight) is represented with the details like GenBank accession numbers, putative function, E-value (BLAST) etc (see Additional data file 3). Homologues of several other proteins, implicated in immunity such as antioxidants, ferritins, transferrins, apolipophorins and apoptotic pathway components were also found, but are not further described here.

The transcriptome under study most likely represents a plethora of E. coli-induced genes in the fat body of the tasar silkmoth. We validated 15 putative immune response genes by semi-quantitative RT-PCR, and their expression profiles are shown in Figure 6. All but two genes, a putative protease inhibitor [GenBank:DQ666519] and a seroin gene [GenBank:DQ666525] were upregulated upon infection. The two DFPs tested, DFP-1 [GenBank:DQ666501] and DFP-4 [GenBank:DQ666503], were induced by both E. coli and M. luteus. DFP-1 was highly expressed in all the tissues with a more prominent expression in mid gut whereas DFP-4 was exclusively expressed in fat body (Figure 6). We analysed the expression pattern of the aforementioned genes in larval tissues differentially challenged with E. coli (gram-negative) or M. luteus (gram-positive) and compared with challenged or mock-challenged (saline-injected) tissues as negative controls. Most of the genes analysed were expressed more prominently upon infection with E. coli than with M. luteus suggesting that there may be differential responses towards different pathogens. Two pathways of differential immune induction have been identified in Drosophila. Gram-positive bacteria and fungi induce the Toll pathway whereas gram-negative bacteria evoke the Imd pathway 37, but the degree of conservation of these pathways between Drosophila and lepidopterans is not known.

A. mylitta, 5^th instar, day 3 larvae were procured from Regional Research Station, Warangal, Andhra Pradesh. Log phase E. coli cells (DH5α), washed and resuspended in saline (0.3 M NaCl, 0.005 M KCl), were injected into the haemocoel of the larvae as described earlier 38. At 24 hours post infection (hpi), larvae were dissected to isolate fat body, and the tissue was flash frozen in liquid N₂ and then stored at -70°C till further use.

Total RNA was extracted from the fat body using Trizol reagent (Invitrogen). The complementary DNA synthesis was carried out using Stratagene ZAP-cDNA^® synthesis kit following manufacturer's instructions. Directional cDNA library was constructed by cloning of cDNA fragments into pBluescript II SK (+) vector and electroporation into E. coli strain DH10B. Insert-containing plasmid clones were sequenced with RV-M primer (5'GAG CGG ATA ACA ATT TCA CAC AGG 3') with the aid of MegaBACE3000 sequencer.

Raw sequences obtained from sequence chromatograms were processed using several programs. A cut-off Phred Quality Value of ≥15 was assigned to extract quality sequences from chromatograms. The quality sequences were screened for the presence of vector sequences using 'Cross Match' program 39. Then masked vector sequences were automatically removed by in-house developed trimming tool. Sequences shorter than 50 bases were removed. The resulting high-quality sequences were assembled into sequence contigs with the TGICL program 40, which initially makes clusters using MegaBLAST and thereafter makes an assembly using CAP3 for each cluster generated in the first step. A cluster is defined as a unique sequence obtained either by multiple alignment of many sequences that are > 95% similar or derived from a single sequence. A cluster containing ≥2 ESTs is termed a contig and that containing only one sequence, a singleton.

The unique putative gene sequences obtained by clustering and assembly were annotated by running BLAST 41 against non-redundant (nr) protein database of NCBI. Further, BLAST output was parsed to classify the putative gene transcripts into different functional classes.

Proteins that did not show any significant hits in NCBI nr database or showed similarity to unknown or hypothetical proteins were characterized by additional computational tools.

A. mylitta larvae were differentially challenged- a) Unchallenged, b) Saline-injected, c) M. luteus and d) E. coli. Log phase E. coli and M. luteus bacteria, washed and resuspended in insect saline (0.3 M NaCl, 0.005 M KCl) were injected (30 μl) into a set of A. mylitta larvae. One set each of saline-injected and uninjected larvae were kept as a control. Four tissues, namely fat body, epidermis, mid gut, and silk gland were dissected out and flash frozen in liquid nitrogen. Total RNA was isolated using Trizol reagent. To remove genomic DNA contamination, total RNA was treated with RNAse free DNAse (NEB) as prescribed by the manufacturers. cDNA was synthesized using MMLV reverse transcriptase (Invitrogen) and oligo dT primers from 1 μg of total RNA. The primers were designed for the selected ESTs by Primer3 software 45.

Semi-quantitative RT-PCR was carried out for all the four differentially challenged tissues using an Eppendorf master cycler under the following conditions- 94°C, 2 min- initial denaturation, 27 cycles (94°C – 30 s, 58°C- 30s, 72°C-2 mins) and a final elongation at 72°C for 10 mins. Actin cDNA was amplified as an endogenous control. PCR reaction components included: 1X buffer, 100μM dNTPs, 1.5 mM MgCl2, 0.5 units Taq polymerase (MBI), 0.5 μM primers. Primer sequences are enlisted in additional data (see Additional data file 5).

We had obtained full-length coding sequences of the DFP-1, 2 and 3 through EST sequencing. To acquire full-length DFP-4 cDNA, we carried out 5' RACE PCR using the 5' RACE kit (Clontech). The 5' ends were amplified by using an adaptor primer and a reverse gene specific primer. PCR was performed for 25 cycles in an Eppendorf master cycler. A 300 bp band was isolated, sequenced and confirmed to be the 5' DFP-4 sequences.