Although Manduca sexta has significantly contributed to our knowledge on a variety of insect physiological processes, the lack of its genome sequence hampers the large-scale gene discovery, transcript profiling, and proteomic analysis in this biochemical model species. Here we report our implementation of the RNA-Seq cDNA sequencing approach based on massively parallel pyrosequencing, which allows us to categorize transcripts based on their relative abundances and to discover process- or tissue-specifically regulated genes simultaneously. We obtained 1,821,652 reads with an average length of 289 bp per read from fat body and hemocytes of naïve and microbe-injected M. sexta larvae. After almost all (92.1%) of these reads were assembled into 19,020 contigs, we identified 528 contigs whose relative abundances increased at least 5- and 8-fold in fat body and hemocytes, respectively, after the microbial challenge. Polypeptides encoded by these contigs include pathogen recognition receptors, extracellular and intracellular signal mediators and regulators, antimicrobial peptides, and proteins with no known sequence but likely participating in defense in novel ways. We also found 250 and 161 contigs that were preferentially expressed in fat body and hemocytes, respectively. Furthermore, we integrated data from our previous study and generated a sequence database to support future gene annotation and proteomic analysis in M. sexta. In summary, we have successfully established a combined approach for gene discovery and expression profiling in organisms lacking known genome sequences.