Domain-directed polymerase chain reaction capable of distinguishing bacterial from host DNA at the single-cell level: characterization of a systematic method to investigate putative bacterial infection in idiopathic disease.

The use of a broad-range PCR to target conserved sequences in the bacterial ribosome has long been recognized as an approach to investigating idiopathic human diseases for the presence of bacteria. An important example is rheumatoid arthritis where the hypothesis of a bacterial etiology remains viable despite failure of previous methods to identify a causative organism. Practical implementation of this strategy, however, was impeded by requirements unique to the study of these diseases. Hence, an adequately characterized method for achieving this has not appeared. We now describe such a method based on the use of a broadly reactive pair of deoxyinosine-containing primers. Detailed characterization addressing these requirements provided evidence that DNA from at least 96% of potential prokaryotic pathogens would be detected. The sensitivity was shown to approximate that of a single organism per reaction. Importantly, this sensitivity was shown to be maintained among multiple targets and to be unimpaired by a large excess of human DNA. Similar results were obtained with extracts of inflamed human synovial fluid to which as little as 0.01 pg of bacterial DNA or, alternately, a single organism per reaction was added. This method was also shown correctly to detect the causative bacteria in clinically infected synovial fluids, further documenting its applicability to such specimens. Finally, it was converted to an in situ method by which bacterial sequences were histochemically localized to Michaelis-Guttman bodies in tissue sections from a patient with malakoplakia, a poorly understood infectious disease. The broad reactivity and high sensitivity achieved by this approach translate into a high likelihood of detecting an unknown bacterium if present in a clinical specimen. Conversely, if properly controlled, the failure of this method to detect such organisms can provide evidence supporting rejection of the bacterial etiology hypothesis, an important aspect unique to this approach. For those bacterial sequences that are detected, the ability to localize them in situ would help define their histologic context and hence facilitate their pathogenetic interpretation. The present method should therefore be appropriate for use in systematically studying rheumatoid arthritis as well as a number of other important idiopathic disorders where a bacterial etiology is suspect.