Optimizing polymerase chain reaction technology for clinical diagnosis.

Several years after widespread recognition of its potential, it appears that polymerase-chain-reaction (PCR) technology is finally approaching licensed application in medical diagnosis. The concept and practice of exponential in vitro amplification of target DNA sequences have profoundly affected almost all aspects of genetic analysis, and in the clinical laboratory have accelerated the trend toward development of specialized units devoted to nucleic-acid-based diagnostics. PCR has been well explored in research settings for the detection of markers of genetic diseases (1) or malignancy, and for diagnosis of infectious diseases. In the latter category, emphasis is being placed on those organisms that are either difficult or slow to culture. Initial licensed diagnostic uses will probably be to detect chlamydia, Lyme disease (borreliosis), and human immunodeficiency virus infection. Other infections that could benefit greatly from rapid, sensitive assays include those due to mycobacteria, mycoplasma, treponemes, and viruses such as hepatitis, papilloma, and herpesviruses. PCR is not the only target amplification technique that has been described, but it is by far the best explored. Practical experience has shown that successful PCR assays require attention to many different phases-specimen preparation, primer selection, amplification conditions, signal authentication, quantitative analysis, and interpretation of clinical significance. Much creativity has gone into optimizing these assays, but the process has taken longer than some had expected, and some unwarranted skepticism for the technology has probably developed as a result. Many specimen-preparation methods can provide representative DNA templates for PCR. The specific choice involves some tradeoff between template quality and technical complexity, with a resulting risk of errors. Traditional extraction procedures, involving combinations of proteases, detergents, denaturants, organic solvents, and high-speed centriftigation, are too complicated for clinical laboratories. This was appreciated long before the advent of PCR, but the extreme sensitivity of this assay has also emphasized the importance of protecting specimens from contamination by stray target sequences during handling. As a result, very minimal but relatively inefficient preparation methods have been tried, such as lysis of cells in hypotonic media, or the direct introduction of unmodified clinical fluid specimens into PCRs. Although these methods may work in some instances, assay sensitivity could be suboptimal because of poor release of amplifiable templates or the presence of inhibitory substances. In this issue, Buffone et al. (2) address the topic of specimen preparation for PCR detection of cytomegalovirus (CMV) in clinical urine specimens. They had earlier reported success in amplifying CMV from the urine of congenitally infected newborns with no specimen pretreatment other than brief centrifugation (3). Because these newborns usually shed very high concentrations of virus, a minuscule sample should suffice. Assay sensitivity for lesser degrees of viruria is uncertain. Buffone et al. found that increasing the volume of urine assayed sometimes resulted in inhibition of the PCR; they now report that this can be avoided by purifying the CMV DNA from urine specimens by binding to glass powder, a method for which commercial kits are available and which appears reasonably practical. The multiple washing steps involved could increase the risk of contamination, and complicate the processing of large numbers of specimens, thus illustrating the tradeoff mentioned earlier. Other processing methods are also worth considering. Viral particles in the urine can be concentrated by ultracentrifugation if suitable equipment is available. Lysis of the pellet is a simple approach that has been used previously (4, 5). Binding of CMV sequences on solid phases coated with capture probes specific for CMV is also appealing, and would be especially helpful when processing types of specimens that contain much extraneous DNA of host or other microbial origin. Eventually, several techniques need to be carefully compared through the use of split specimens, with special consideration being given to methods that can be automated. Choosing appropriate primers and targets for diagnostic PCR has been largely empirical. For microbial DNA targets, primer sequences should be conserved in clinical isolates, yet be specific to the desired target and organism. Numerous parts of the CMV 230-kb genome are suitable for amplification. Many investigators, including Buffone et al. (2,3), have used primers from its major immediate-early (WE) gene region. Published data from two reference strains of CMV show fairly good sequence conservation in this region, although additional data from our laboratory reveal nucleotide variation in clinical strains that affects two of the four MIE primers listed (2,3; MIE-5 and CMV18 are variant; the sequence of CMV24 should be verified). As the data of Buffone et al. show, some primer-template mismatching does not necessarily prevent amplification of various strains. However, some mismatch could affect assay sensitivity or the ability to obtain representative amplification of the sequences present in the specimen. It is best to assess the potential severity of the problem by sequencing diverse strains and attempting to amplify variants of known sequence with the chosen primers. Variable genomes such as human immunodeficiency virus need to be screened carefully to select relatively conserved sequences. It may be necessary to use more than one set of primers.