DNA microarray technology has been extensively used for gene expression analysis of both eukaryotic and prokaryotic organisms. For eukaryotic gene expression profiling, the poly(A)-based reverse transcription of messenger RNA (mRNA) followed by T7 RNA polymerase-based in vitro transcription is generally required to produce enough RNA targets for hybridization with the microarray chips. However, the same method cannot be directly applied to prokaryotic mRNAs due to the lack of poly(A) sequences at the 3' ends. Conventional methods usually require large amounts of starting RNAs and lead to high background noise. Recently developed amplification methods enable smaller amounts of prokaryotic RNA to be used from samples with species-specific primers, oligo(dT) primers, or random primers. In this study, three target preparation methods, including the direct labeling, polyadenylation-involved oligo-dT priming, and random priming amplification (respectively referred to as DL, PAOD, and RPA hereafter) were evaluated through expression profiling of a heat shock model of Escherichia coli. The PAOD method was found to be more sensitive and more specific in differential gene expression measurements than either DL and RPA, even when the E. coli RNA was only a small proportion of the simulated eukaryotic host RNA. The results suggest that PAOD is the preferred target preparation method for prokaryotic transcriptome.