This study of pediatric RTI used a DNA microarray that aims to detect all known viral species simultaneously. The most common viral pathogens identified were RSV (19%), picornaviruses (16%), and influenza A/B (11%). Notably, most viruses detected in the respiratory tracts of patients with non-respiratory illnesses and in immunocompromised patients were picornaviruses. These findings are consistent with the observation that asymptomatic rhinovirus infections can be seen in 4 – 12% of healthy individuals and suggest that immunocompromised hosts may be more susceptible to colonization or overt respiratory illness by picornaviruses (23
). Our frequency of detection of 16% for picornaviruses is consistent with that of a previously published report (18% rhinoviruses, 2.9% enteroviruses), which also used consecutive NPA samples (24
Our comparison of the Virochip with DFA demonstrates that the sensitivity of the microarray for respiratory virus detection is superior. Importantly, about 50% of the overall increase in detection rate corresponds to samples with inconclusive DFA results due to low cellular content. Unlike DFA, nucleic-acid detection methods such as microarray and PCR are capable of detecting free viral particles in addition to virus-infected cells. Another significant advantage of a pan-viral microarray over DFA is the ability to screen for all known viruses simultaneously. DFA panels in current clinical use do not test for RV/EV, HMPV, or CoV, which, in our study, comprised more than one-third of the detected viruses in patients with RTI.
In the 17 patients presenting with an illness severe enough to require mechanical ventilation, there were three cases of critical RV infection, including one in an immunocompromised individual. RV infection is thought to comprise a spectrum of disease ranging from asymptomatic infection to life-threatening childhood pneumonia (25
). Our findings of cases of critical respiratory tract illness associated with RV infection is consistent with growing evidence linking RV with hospitalizations in young children (26
). The Virochip was also superior to DFA for detecting double viral infections. Two cases of double infection, in which both viruses could in principle be detected by DFA, were reported as single-virus infections by DFA. Previous studies have suggested that double infections are associated with greater severity of RTI (27
). Higher efficiency of the microarray in detecting critical as well as double viral infections is an important advantage of the method, as timely detection of such infections may allow clinicians to avoid unnecessary antibiotics and invasive procedures and begin appropriate antiviral treatment, if available.
In addition to being a highly parallel methodology that utilizes thousands of oligonucleotide probes for simultaneous detection, a DNA microarray platform Virochip is expandable and adaptable. Automated oligonucleotide design methods allow straightforward addition of new probes for better detection of known viruses or to expand coverage to novel or evolving viral species (16
). DNA microarrays also can be used for non-viral pathogen detection, including bacteria and fungi (30
). Interestingly, in this study the Virochip detected Streptococcus pyogenes
bacteriophage in a sample from a patient with aspiration pneumonia who also had a positive sputum culture for Streptococcus pyogenes
. This result suggests a possible strategy of using phage sequences as an indirect means of detecting bacterial pathogens.
Despite its pan-virus scope, the Virochip only detected a virus in 64% of RTI cases overall. Although comparable with studies using other detection methods (1
), this result is likely an underestimate of the true number of positives arising from several factors. First, some samples may have been missed by the Virochip due to low virus titers at the detection limits of PCR. Second, RNA (not DNA) was used as the source material in this study, which may have given rise to lower than expected rates of detection of DNA viruses such as adenoviruses, herpesviruses, and parvoviruses. The Virochip may not have detected any cases of human bocavirus, a recently described parvovirus (31
), because this virus shares low sequence identity to parvovirus sequences currently represented on the Virochip. Finally, 32% of patients with RTI in this study were diagnosed with pneumonia, of which 35% were hospital-associated. Many of these cases of pneumonia may be non-viral in origin, as suggested by previous epidemiologic studies of pneumonia in hospitalized children (32
Using custom PCR primers, we confirmed at least three of the six microarray-positive/PCR-negative cases (one RSV, one RV, and one EV) as true positive tests for viruses. Detection of such cases by microarray is not unexpected, as the Virochip uses multiple probes that are derived from different sites in the viral genome and have a higher tolerance for sequence mismatches than the primers used in specific PCR. This results in improved detection of divergent viruses (14
). We suspect that of the microarray-positive/PCR-negative cases of virus infection likely were missed by clinical PCR due to mismatches between primer and target sequences (35
), although specimen availability limited our efforts to show this definitively.
Although the prospect of comprehensive viral detection by use of a single microarray assay is appealing, several challenges must be addressed before use of the technology in clinical diagnosis becomes practical. First, there are significant initial start-up costs in setting up the technology including the cost of the microarray printer and viral oligonucleotides, as well as ongoing material and labor costs. Second, the turnaround time for the assay is currently ~24 hours; it may be possible to decrease the processing duration by using ultra-rapid polymerases for amplification or controlled agitation techniques during hybridization. Third, reproducibility and consistent array quality are of concern. In our study, none of the microarray assays needed to be repeated, and the inherent probe redundancy built into the method makes the assay robust for purposes of virus detection. For example, of the 300 oligonucleotide probes on the Virochip designed to hybridize to rhinoviruses, as few as four high intensity oligonucleotides are sufficient to make a successful virus identification. Statistical methods for interpreting the microarray data, as exemplified by E-Predict (17
), can be completely automated, allowing ease of use and freedom from operator bias. Design of smaller microarrays aimed at detection of targeted virus sets (e.g.
respiratory viruses only) would reduce cost and simplify issues of reproducibility, quality control, and data analysis.
Looking to the future, we can envision two possible ways in which a viral detection microarray could be used to impact clinical practice. One strategy is to use the platform for direct diagnosis of respiratory infections, as reported here. An alternative strategy may be to deploy a viral microarray as an instrument of discovery rather than routine detection, with the goal of identifying divergent or unexpected viruses that elude diagnosis using conventional methods. Once a candidate new pathogen is identified, a specific PCR-based or DFA-based assay then can be developed to detect the virus with a high degree of sensitivity in clinical samples. In this scenario, microarray assay would complement rather than replace existing diagnostic techniques such as PCR.