Transcriptional approaches are increasingly used to compare the behaviour of pathogenic and non-pathogenic bacteria in different culture conditions. The purpose of this study was to apply these methods in cheese to better characterize food and clinical Enterococcus faecalis isolates during cheese processing. Because of the complex biochemical composition of the cheese matrix, e.g. the presence of casein and fat, we developed an efficient method to recover total RNA from bacteria in a semi-hard cheese model. To validate the RNA extraction method, we analysed expression of 7 genes from two E. faecalis strains (one clinical and one food isolate) in both cheese and culture medium by semi-quantitative RT-PCR. We then used PCR-based DNA macro-arrays to compare expression of 154 genes from two E. faecalis strains in both cheese and culture medium. The food strain isolated from cheese is transcriptionally active in cheese, as reflected by the higher transcript levels of various genes. Conversely, overall transcript levels of the V583 clinical isolate were lower in cheese, suggesting that the food strain may be more adapted to a dairy environment than the clinical strain. The method described here constitutes a very promising tool for future transcriptomic studies in cheese matrices. Global profiling in foods may prove to be a valid criterion for differentiating food from clinical isolates.