Multiplexed Reverse Transcriptase PCR Assay for Identification of Viral Respiratory Pathogens at the Point of Care

Abstract

We have developed a nucleic acid-based assay that is rapid, sensitive, and specific and can be used for the simultaneous detection of five common human respiratory pathogens, including influenza virus A, influenza virus B, parainfluenza virus types 1 and 3, respiratory syncytial virus (RSV), and adenovirus groups B, C, and E. Typically, diagnosis on an unextracted clinical sample can be provided in less than 3 h, including sample collection, preparation, and processing, as well as data analysis. Such a multiplexed panel would enable rapid broad-spectrum pathogen testing on nasal swabs and therefore allow implementation of infection control measures and the timely administration of antiviral therapies. We present here a summary of the assay performance in terms of sensitivity and specificity. The limits of detection are provided for each targeted respiratory pathogen, and result comparisons were performed on clinical samples, our goal being to compare the sensitivity and specificity of the multiplexed assay to the combination of immunofluorescence and shell vial culture currently implemented at the University of California-Davis Medical Center hospital. Overall, the use of the multiplexed reverse transcription-PCR assay reduced the rate of false-negative results by 4% and reduced the rate of false-positive results by up to 10%. The assay correctly identified 99.3% of the clinical negatives and 97% of the adenovirus, 95% of the RSV, 92% of the influenza virus B, and 77% of the influenza virus A samples without any extraction performed on the clinical samples. The data also showed that extraction will be needed for parainfluenza virus, which was only identified correctly 24% of the time on unextracted samples.

Each year, between October and March, hospital admissions suddenly increase with patients presenting with influenza or influenza-like symptoms. It is estimated that influenza-associated hospitalizations in the United States range from approximately 54,000 to 430,000 per season (5). Respiratory syncytial virus (RSV) is the most common cause of bronchiolitis and pneumonia among infants and children under 1 year of age, but most respiratory viruses can trigger severe lower respiratory tract disease at any age, especially among the elderly or among those with compromised cardiac, pulmonary, or immune systems (6). In this context, timely and accurate identification of respiratory viruses is rapidly becoming more relevant as antiviral treatment options increase. In addition, the resulting improved treatment of patients presenting with respiratory illness will help control infection, prevent nosocomial spread, and reduce patient stay as well as hospital costs.

Although alternative respiratory virus identification techniques such as immunofluorescence and rapid antigen detection tests have been developed to provide rapid diagnostic capabilities, viral culture remains the most prevalent test in use for laboratory identification (15). The main drawback of immunofluorescence and rapid test kits is their lack of sensitivity. A recent study reported that immunofluorescence assays detect only 19% of respiratory viruses with viral loads below 10⁶ copies/ml (9), and rapid test kits have been shown to have typical false-negative rates of 30% for influenza virus (13). Although viral culture is both sensitive and specific, it is labor-intensive and time-consuming. In addition, because some viral strains grow poorly and/or slowly in cell culture, timely results are not available to impact or inform clinical decisions such as the use of antiviral drug treatment. A recent study undertaken with pediatric patients to determine the impact of rapid diagnosis of influenza (such as the FluOIA test from Biostar, Inc., which is only 83 to 96% sensitive and 64 to 76% specific) on physician decision-making and patient management in the emergency room showed that the use of rapid test kits at point of care lead to a reduction of antibiotic prescriptions of 40%, a reduction of laboratory and radiograph charges of 50%, patient discharges occurring 1 h more quickly, and an increase in antiviral use by 25% (4). Another study comparing cell culture and immufluorescence focused on the benefits of rapid reporting of respiratory viruses concluded that the mean length of stay for hospital inpatients with respiratory viral isolates was 10.6 days (mean cost of $7,893) when the patients were diagnosed by viral culture and only 5.3 days (mean cost of $2,177) when they were diagnosed using immunofluorescence (1).

To alleviate issues of specificity and sensitivity inherent to the rapid tests, as well as the long turnaround times of viral culture, laboratories analyzing clinical samples are progressively moving toward molecular diagnostics as a mean to identify respiratory viruses. Nucleic acid amplification techniques such as PCR, followed by gel electrophoresis (7), and quantitative PCR (q-PCR) with corresponding probes (14, 16) have recently been developed for the rapid detection of respiratory pathogens, leading to significant sensitivity and specificity improvements over culture and immunofluorescence techniques. Nevertheless, a limitation of semiquantitative real-time PCR assays is their extremely low level of multiplexing. Multiplexed detection capabilities provide many advantages over conventional detection methodologies. In the event of a respiratory disease outbreak, the use of multiplexed assay panels can provide a cost-effective means of handling high volumes (i.e., a surge) of samples. Moreover, custom tailored assay panels designed to respond to genetic mutations and/or new pathogens can be rapidly implemented and therefore greatly help reduce the impact of infectious disease outbreaks. In addition, in contrast to current q-PCR assays, which require DNA/RNA extraction, the only requirement of our assay is a nasal swab in buffer solution, dramatically reducing processing time and reagent costs.

We have extended the utility of nucleic acid amplification techniques by developing a multiplexed reverse transcriptase PCR (RT-PCR) assay that allows the timely simultaneous detection of five respiratory viruses. The multiplexed assays (liquid arrays) have been developed on a commercially available flow cytometer (Bioplex; Bio-Rad, Inc.). The assay utilizes surface-functionalized polystyrene microbeads, embedded with precise ratios of red and infrared fluorescent dyes (Fig. 1). There are 100 unique dye ratios, giving rise to 100 unique bead classes. When excited by a 635-nm laser, the two dyes emit light at different wavelengths (658 and 712 nm), and thus each bead class has a unique spectral address. Bead classes can be easily distinguished; therefore, they can be combined, and up to 100 different analytes can be measured simultaneously within the same sample. Although liquid arrays have been demonstrated in a variety of applications (8), including the detection of antigens, antibodies, small molecules, and peptides, in the currently described application, beads are functionalized with a nucleic acid probe approximately 30 bases long, where the probe sequence is complementary to a target amplicon. Nucleic acid from the pathogen of interest is amplified by RT-PCR (Fig. 1), which is conducted using a mixture of all forward and reverse primers for each of the pathogen targets in the multiplexed panel. The amplified product is then introduced to the bead mixture, allowed to hybridize, and subsequently labeled with the fluorescent reporter, streptavidin-phycoerythrin (SAPE). Each optically encoded and fluorescence-labeled microbead is evaluated with the Bioplex flow cytometer. A red laser excites the dye molecules inside the bead and classifies it, while a green laser quantifies the assay at the bead surface via the median fluorescence intensity (MFI) of the SAPE reporter. The flow cytometer is capable of reading several hundred beads each second, and fluorescence analysis can be completed in as little as 15 s.

FIG. 1.

Individual primer pairs (biotinylated forward and standard reverse) that bracket the targeted genomic sequence are included in an RT-PCR master mix. After target amplification by RT-PCR, the amplicons are mixed with beads and target amplicons containing the forward biotinylated primer hybridize to the complementary probe on the appropriate beads. A fluorescent reporter molecule (streptavidin-phycoerythrin) then binds biotin functional groups. The completed assay includes a bead, a probe, and a biotinylated and fluorescently tagged amplicon. The sample is then analyzed by using a flow cytometer, and an MFI value is reported for each bead class, with each bead class representing a specific signature.

The current panel (Table 1) includes 16 beads, with assays for influenza virus A (two assays); influenza virus B (two assays); parainfluenza virus types 1 (one assay) and 3 (one assay); RSV (one assay); and adenovirus groups B (two assays), C (two assays), and E (one assay). The panel also includes four unique internal controls described in Materials and Methods. Typically, results on a clinical sample can be provided in less than 3 h, including sample collection, preparation, and processing, as well as data analysis.

TABLE 1.

Summary table of the 16-plex respiratory panel layout^a

Target	Signature	Sequence
		Biotinylated forward primer	Reverse primer	Probe
Virus
Influenza virus A	Flu A-1	Available upon requestb	Available upon requestb	Available upon requestb
	Flu A-2	5′/5Bio/GGACC/iBiodT/CCACTTAC/iBiodT/CCAAAACAGAAAC-3′	5′/GTAAGGCTTGCATGAATGTTATTTGCTC-3′	5′/5AmMC6//iSp18/TTGACCTAGTTGTTCTCGCCA-3′
Influenza virus B	Flu B-1	Available upon requestb	Available upon requestb	Available upon requestb
	Flu B-2	5′/5Bio/GTCCA/iBiodT/CAAGCTCCAG/iBiodT/TTT-3′	5′/TCTTCTTACAGCTTGCTTGC-3′	5′/5AmMC6//iSp18/CCTCCGTCTCCACCTACTTCGTT-3′
RSV	RSV	Available upon requestb	Available upon requestb	Available upon requestb
Para-influenza virus 1	Para 1	5′/5Bio/ATGCTCC/iBiodT/TGCCCACTG/iBiodT/GAATG-3′	5′/AATCTTTATCCCACTTCCTACACTTG-3′	5′/5AmMC6//iSp18/TCTATACCTTCACTCGAGTAATCTG-3′
Para-influenza virus 3	Para 3	5′/5Bio/ACCAGGAAAC/iBiodT/ATGC/iBiodT/GCAGAACGGC-3′	5′/GATCCACTGTGTCACCGCTCAATACC-3′	5′/5AmMC6//iSp18/AGAGCTCCTAAACATGATGGATACC-3′
Adenovirus B	Adeno B-1	5′/5Bio/TCCTGCACCA/iBiodT/TCCCAGA/iBiodT/A-3′	5′/CCTCCGGGACCTGTTTGTAA-3′	5′/5AmMC6//iSp18/CTGACACGAATAATTCAAGGCTGGAAAGCTG-3′
	Adeno B-2	5′/5Bio/CGCTT/iBiodT/CACAGTCCAAC/iBiodT/GC-3′	5′/GCTGCTTGTGGGTTTGATGA-3′	5′/5AmMC6//iSp18/CGTTTTCGGATTATGATTCCCATCGTTCTTC-3′
Adenovirus C	Adeno C-1	5′/5Bio/AGCGCG/iBiodT/AATATTTGTC/iBiodT/AGGGC-3′	5′/TCAGCTGACTATAATAATAAAACGCCA-3′	5′/5AmMC6//iSp18/CGGAACGCGGAAAACACCTGAGAAAA-3′
	Adeno C-2	5′/5Bio/TCGA/iBiodT/CTTACC/iBiodT/GCCACGAG-3′	5′-GCCACAGGTCCTCATATAGCAA-3′	5′/5AmMC6//iSp18/TGCTCCACATAATCTAACACAAACTCCTCACCC-3′
Adenovirus E	Adeno E	5′/5Bio/TGCAAT/iBiodT/TTGTTGGGT/iBiodT/TCG-3′	5′/CCTGGCTGTTATTTTCCACCA-3′	5′/5AmMC6//iSp18/TTAATCATGGTTCTTCCTGTTCTTCCCTCCC-3′
Controls
RNase P	RNase P	5′/5Bio/AGA T/iBiodT/T GGA CC/iBiodT/GCG AGC G-3′	5′/GAGCGGCTGTCTCCACAAGT-3′	5′/5AmMC6//iSp18/TTC TGA CCT GAA GGC TCT GCG CG-3′
Mt7	Mt7	NA		5′/5AmMC6//iSp18/CAAAGTGGGAGACGTCGTTG-3′
Mt7-Cy3	Mt7-Cy3	NA		5′/5AmMC6//iSp18/CAAAGTGGGAGACGTCGTTG-3′Cy3
Mt7-biotin	Mt7-biotin	NA		5′/5AmMC6//iSp18/CAAAG/iBiodT/GGGAGACGTCG/iBiodT/TG-3′

^aThe biotinylated forward (“Bio” denotes a biotin placed at the 5′ end, while “iBiod” denotes an internal biotin) and reverse primer sequences are provided for each signature contained in the panel. The probe design is also detailed, including the 5′-end reactive group (AmMC6, amino modifier C6, also called phosphoramidite) and the spacer 18 (noted as “iSp18”), which is an 18-atom hexa-ethyleneglycol spacer placed between the reactive group and the DNA sequence to allow optimal coupling of the carboxylated bead to the probe.

^bUnpublished sequence, provided to Lawrence Livermore National Laboratory by the CDC and available upon request. Contact Stephen Lindstrom (e-mail: sq15@cdc.gov; phone: [404] 639-1587; fax: [404] 639-0080) for information regarding influenza signatures and Dean Erdman (e-mail: ddel@cdc.gov; phone: [404] 639-3727; fax: [404] 639-4416) for information regarding RSV signatures.

This article presents a summary of the assay performance in terms of sensitivity and specificity. Limit-of-detection (LOD) values for each targeted respiratory pathogen are presented for the multiplexed panel, and result comparisons are performed on clinical samples collected at the University of California Davis Medical Center, Davis, CA (UCDMC), our goal being to compare the sensitivity and specificity of the multiplexed assay to the currently implemented detection techniques.

MATERIALS AND METHODS

Reagents.

Tris-NaCl (0.1 M Tris, 0.2 M NaCl, 0.05% Triton X-100, [pH 8.0]) and TE (10 mM Tris-HCl, 1.0 mM EDTA [pH 8.0]) buffers were purchased from Teknova, Inc. (Hollister, CA). SAPE was purchased from Invitrogen, Inc. (Carlsbad, CA) and suspended in Tris-NaCl at a concentration of 3 ng/μl. All primers and probes were synthesized by Integrated DNA Technologies (Coralville, IA) and suspended in TE buffer.

Viruses.

Current circulating strains of certified killed respiratory viruses were purchased at a stock concentration of 1 mg/ml. Influenza A viruses (A/H1, New Caledonia strain, and A/H3, Shandong strain), RSV, and adenovirus C were purchased from Research Diagnostics, Inc. (Flanders, NJ), while influenza virus B, Victoria strain, and parainfluenza virus types 1 and 3 were purchased from Advanced Immunochemical, Inc. (Long Beach, CA). Adenovirus group B and E strains were also grown, and titers were determined by the method of Reed and Muench (12).

Carbodiimide coupling of amino-substituted probes to carboxylated microbeads.

Different sets of carboxylated fluorescent microbeads were obtained from Luminex Corp. (Austin, TX), and oligonucleotide probes for the respiratory panel were assigned to individual bead sets. Each probe sequence represented the reverse complement to the target region of the forward strand (5′-3′) and contained a spacer (18-atom hexa-ethyleneglycol spacer) between the reactive group (amino modifier C6, also called phosphoramidite) and the 5′ end of the oligonucleotide to enable optimal hybridization. Phosphoramidite is a primary amine that results in a stable, covalent attachment upon reaction with the ester on the bead coating. Probes for each of the pathogen targets were coupled to the beads according to the manufacturer's recommended coupling protocol. Briefly, a 1-ml aliquot of beads (1.25 × 10⁷ beads) was resuspended in 50 μl of 0.1 M 2-[N-morpholino] ethanesulfonic acid (MES) buffer at pH 4.5 and sonicated. Then, 0.05 mg of 1-ethyl-3-[3-dimethylaminopropyl]-carbodiimide hydrochloride (EDC; Pierce Biotechnology, Rockford, IL) was added, along with 10 μl of probe at a concentration of 50 μM. This solution was incubated in the dark at room temperature for 30 min. A second aliquot of EDC (0.025 mg) was added, followed by incubation under the conditions described above. The beads were then rinsed in 1 ml of phosphate-buffered saline containing 0.02% Tween 20 (Sigma, St. Louis, MO), centrifuged at 10,000 rpm for 5 min, rinsed in phosphate-buffered saline containing 0.1% sodium dodecyl sulfate, centrifuged a second time, resuspended in 250 μl of TE buffer, sonicated, and stored in the dark at 4°C. A 10× bead set containing all conjugates was then prepared, using 200 μl of each bead in a total volume of 5 ml of Tris-NaCl buffer. A 1× working solution was then prepared from the stock before use, using a Tris-NaCl buffer for dilution.

RT-PCR.

All RT-PCR reactions were prepared by using the end-point Superscript III One-Step RT-PCR kit from Invitrogen, Inc. Typically, each 25-μl PCR contained 12.5 μl of Superscript III Master mix, 0.5 μl of MgSO₄ (50 mM), 0.1 μl of each forward and reverse primer (0.4 uM final concentration), 1 μl of reverse transcriptase and Taq DNA polymerase mix, and PCR-grade water to complete the volume to 20 μl. A total of 5 μl of unextracted sample was then added to 20 μl of PCR mix, and the mixture was cycled on a thermocycler using the following parameters: reverse transcription at 50°C for 30 min, denaturation at 95°C for 15 min, and then 35 PCR amplification cycles of denaturation at 94°C for 15 s, annealing at 55°C for 30 s, and extension at 72°C for 15 s.

Microbead hybridization.

After RT-PCR, 5 μl of amplified product was added to 22 μl of bead mix and hybridized to the probe-coated beads by using a denaturation step at 95°C for 2 min, followed by an annealing step at 55°C for 5 min.

Microbead washing and labeling.

The hybridized bead solution was transferred to a 96-well filter plate (Millipore, Inc., Bedford, MA) with 1.2-μm pores. The beads were washed three times to remove unbound oligonucleotides by using 100-μl aliquots of Tris-NaCl buffer pipetted in each well and then vacuum aspirated with a vacuum manifold kit (Millipore). The washed beads were next incubated with 60 μl of 3 ng of SAPE reporter/ml for 5 min, washed twice with 100-μl aliquots of Tris-NaCl buffer, and transferred into a 96-well microtiter round bottom plate. For each well, 50 μl of solution was analyzed in the Bioplex flow analyzer.

Controls.

Controls that convey important diagnostic information regarding reagent addition, quality, and concentration, assay operator performance, and instrument stability were added to the assay. A unique set of four internal controls are built into every sample, monitoring and reporting every step of the protocol. The negative control (NC) is a bead coupled to a Mt7 probe. Mt7 is a nucleic acid sequence obtained from Thermotoga maritima, an organism found near deep-sea thermal vents. This organism was selected to serve as a NC because its nucleic acid is unlikely to be observed in clinical samples. Thus, Mt7 is not expected to bind exogeneous nucleic acids and, consequently, the median fluorescence intensity (MFI) of the NC beads should always be low. High MFIs on the NC beads obtained in the presence of a sample would indicate a lack of specificity. The instrument control (IC) verifies the reporter fluorescence optics of the flow analyzer. The IC is a bead to which a Cy3-labeled Mt7 probe has been coupled. The probe is unlikely to bind other nucleic acids, and the Cy3 dye emits a constant fluorescence (i.e., constant MFI) in all samples when excited by the reporter laser. A change in MFI on the IC bead indicates fluctuations in the reporter laser performance. The fluorescent control (FC) tests for the addition of the fluorescent reporter (SAPE). FC is a bead coupled with biotinylated Mt7 probe that fluoresces after exposure to SAPE. A bead coupled to an RNase P probe serves as a positive PCR control, as well as a control for the addition of the clinical sample. Signals are obtained only when PCR product has been generated and bound to the probe, and SAPE has been added; lack of signal on the PCR control bead indicates that either PCR was not performed properly or that SAPE was not added. The FC control, however, will yield a signal even in the absence of PCR, so these two events can be decoupled. These controls afford high confidence that the assay is performed correctly by monitoring the addition of sample, confirming PCR was performed, indicating that SAPE was added, checking that the instrument is performing correctly, and verifying that the assay is specific. Every sample was analyzed in the context of the performance of the controls, thereby minimizing the likelihood of false-positive results.

LOD data.

Each virus was diluted in distilled water starting from a 10²-ng/μl stock. The concentration range for the LOD study spanned 10 orders of magnitude using two dilutions per order of magnitude. Each concentration was run in quadruplicate, and LOD data sets for each specific virus were run on separate 96-well plates in order to prevent any possible cross-contamination. All experiments were performed on whole virus without nucleic acid extraction. Each plate contained eight blank wells in which distilled water was added to the RT-PCR mix as negative controls.

Clinical sample collection and handling.

From November 2004 through November 2006, more than 1,000 nasal swab samples were collected from patients arriving in the emergency room at the UCDMC Emergency Department in Sacramento, CA, which treats 60,000 patients per year, including 12,000 children. Nasal swabs were obtained from patients showing respiratory symptoms, as well as from asymptomatic subjects such as accompanying family members.

Nasal swabs were collected in 3 ml of M4 viral transport medium (Remel, Lenexa, KS), which is composed of gelatin, vancomycin, amphotericin B, and colistin. The sample was then deidentified and divided into two tubes. One aliquot was subjected to immunofluorescence testing and/or viral culture by utilizing a standard shell vial technique, while the other sample aliquots were analyzed with multiplexed assays on the Bioplex platform. According to the immunofluorescence and/or viral culture results, the clinical sample inventory contained 56 RSV samples, 35 influenza virus A samples, 12 influenza virus B samples, 46 parainfluenza virus samples, 30 adenovirus samples, and 828 negative samples.

Extraction.

Although extraction was not generally performed on clinical samples, results obtained with parainfluenza virus were suboptimal. In order to asses whether these results derived from poor primer performance or had other roots, viral RNA was purified for eight parainfluenza virus samples by using the MagMAX-96 viral RNA isolation kit (Ambion 1836). During the purification process, the samples were lysed, and magnetic beads were used to bind the nucleic acid. The beads were then washed using two alcohol wash solutions. After the washes, the nucleic acid was removed from the beads by adding an elution buffer and heating the solution to 65°C. This eluent, which represents the purified RNA sample, was used for the multiplexed RT-PCR.

RESULTS

Respiratory panel design.

An initial set of 24 signatures derived from a variety of sources (Centers for Disease Control, Lawrence Livermore National Laboratory Bioinformatics Group, as well as previously published work [14, 16]) were chosen for their ability to bind and amplify target-specific genes which are phylogenetically conserved and therefore insensitive to strain variations. Each signature typically consisted of two 20-bp primers and a 30-bp probe; the typical amplicon length was 90 to 200 bp.

Signatures were first tested in singleplex reactions (with only one primer pair present in the primer mix) against their respective targets in order to ascertain the likelihood of identifying target. Four signatures did not identify target except at very high concentrations (100 pg per reaction) and were therefore discarded, leaving 20 signatures for assembly in a multiplexed panel. Starting with a single viral target, individual signatures were added to a growing mixture one at a time, until all target signatures were added and demonstrated to work as effectively in the multiplexed environment as they did in the singleplex format. This viral target signature “block” was then combined with another viral signature block and tested again. After each signature addition, poor performers and/or competing signatures were isolated and removed. Poor performers were typically signatures that provided low but adequate MFI signals in singleplex but for which the MFI signals further dropped in the presence of other signatures. Competing signatures were comprised of primer sets that amplified overlapping target regions and therefore competed for target amplification. The effect of such a competition is a concomitant drop of the MFI signal for both competing signatures while other signatures keep performing well in the assay.

Assay optimization.

At the end of this iterative process, the multiplexed respiratory panel was composed of twelve signatures and four controls. RT-PCR parameters such as the added MgSO₄ concentration, the annealing temperature, the extension temperature, and the extension time, were then optimized for this final respiratory panel in order to produce a combination of low backgrounds, high MFI signals, and low cross-reactivity. Four MgSO₄ concentrations ranging from 1 to 6 mM, three annealing temperatures ranging between 50 and 60°C, two extension temperatures (68 and 72°C), and three extension times ranging from 10 to 20 s were investigated. For each new parameter under study, an LOD curve was built in triplicate for a minimum of three organisms, including adenovirus type C, RSV, and influenza virus A, and the experimental conditions leading to the best combination of background, MFI, and cross-reactivity signals across the range of targets tested was selected (data not shown). All of the primer concentrations were maintained at 0.4 μM, except for the RNase P control primer concentration, which was decreased to 0.2 μM to reduce the probability of amplification competition, and the RSV forward primer concentration, which was increased to 0.8 μM due to the fact that two reverse primers are present in the mix, amplifying RSV types A and B, respectively. Details of the optimized RT-PCR protocol are provided in Materials and Methods.

LODs.

The LOD for each target was then determined with the 16-plex respiratory panel, using the protocol described in Materials and Methods. An average of the four MFI values was plotted on the LOD graph for each concentration, as well as the standard deviation. Two examples are provided in Fig. 2: the LOD curves for influenza virus A are shown in Fig. 2A for both influenza virus A signatures when titrating using the New Caledonia strain, which is an A/H1 subtype. The LOD curve for the single signature for parainfluenza virus 3 is also presented in Fig. 2B. A summary of the LOD values, defined as virus concentrations at which the corresponding average MFI values were above the background by more than three standard deviations, is presented in Table 2. The LOD value obtained for parainfluenza virus 1 was higher than for the other viruses. As discussed in the clinical evaluation section below, this result was attributed to the remarkable stability of the nucleocapsid that encapsidates the RNA of paramyxoviruses (17). All of the other LOD values obtained with the multiplexed RT-PCR assay without performing any RNA/DNA extraction step were within 1 to 2 orders of magnitude of the LOD values published using both RNA/DNA extraction procedures and significantly lower levels of multiplexing (3, 10, 11, 18). The ability to remove the extraction step from the assay protocol may be valuable for point-of-care applications because it simplifies the handling of clinical samples, lowers the processing costs, shortens the analysis time by up to 30 min, and allows for easier assay automation.

FIG. 2.

LOD determination for influenza virus A (two signatures) (A) and parainfluenza virus (one signature) (B). The MFI signals from the other 14 bead classes corresponding to the 14 additional target analytes, as well as the four controls, have been omitted for clarity.

TABLE 2.

Summary table of LOD values in the multiplexed respiratory panel^a

Signatureb	LOD in multiplexed RT-PCR panel
Flu A (pg/μl)	0.005
Flu B (pg/μl)	0.01
Para 1 (pg/μl)	5,000
Para 3 (pg/μl)	0.05
RSV (pg/μl)	0.1
Adeno B* (TCID50/μl)	0.1
Adeno C (pg/μl)	0.005
Adeno E* (TCID50/μl)	0.001

^aLOD values for each targeted respiratory virus in the multiplexed respiratory panel are presented. The LOD value was defined as the virus concentration at which the corresponding average MFI value was above the background by more than three standard deviations.

^b*, only available as live virus.

Clinical evaluation.

The multiplexed panel was tested on clinical samples collected from patients arriving in the emergency room at the UCDMC. Nasal swabs were collected in viral transport medium and divided into two aliquots. One aliquot was diagnosed by using immunofluorescence and/or viral culture, while the other aliquot was diagnosed by using the multiplexed RT-PCR respiratory panel on the Bioplex platform. For the Bioplex-based assay, 5 μl of nasal swab sample was directly mixed with 20 μl of PCR reagents, and the amplification, bead hybridization, washing, labeling, and flow cytometer analysis steps were performed according to the previously described protocol (see Materials and Methods details). A total of 828 negative samples were first analyzed in order to set threshold values for positive identification. Threshold values for each signature were calculated based on the response of the known negative patient samples. First, outliers were removed iteratively using the Grubb's outlier test (2). After the outliers were removed, thresholds were calculated for each signature. The threshold value was chosen such that the MFI values of negative samples that were not determined to be outliers would exceed this value at a rate of 0.005, which corresponds to a set assay specificity of 99.5%. These thresholds led to a rating scale for which MFI values below the threshold were ruled negative, and MFI values equal to or above the threshold were ruled positive. A summary of the threshold values is provided in Table 3. For viruses for which two signatures were included in the panel, a positive was called when at least one of the signatures had an MFI equal to or above threshold. Since threshold values are determined from the background response, which is dependent on the sample matrix, the values listed in Table 3 would remain valid for any nasopharyngeal swab collected according to the protocol described in Materials and Methods. Clinical samples collected in a different manner might present a different matrix composition and thus have a different background MFI response, which would require a separate threshold determination. Of the 828 samples tested, 791 were confirmed to be negative by multiplexed RT-PCR (95.5%) and 37 were identified as positive for a respiratory virus. These 37 samples were sent to the Viral and Rickettsial Disease Laboratory (VRDL) at the State of California Health and Human Services Agency (Richmond, CA) for third party confirmatory q-PCR analysis. The positive multiplexed RT-PCR result was validated for 31 samples and invalidated for 6 samples, bringing the percentage of correctly identified clinical negatives to 99.3% and reducing the rate of false negatives by 4% compared to the combination of immunofluorescence and/or shell vial culture implemented at the UCDMC.

TABLE 3.

MFI thresholds for positive sample identification^a

Signature	MFI threshold for “positive” identification
Flu A-1	112
Flu A-2	40
Flu B-1	63
Flu B-2	119
RSV	181
Para 1	8
Para 3	25.5
Adeno B-1	27
Adeno B-2	47
Adeno C-1	59
Adeno C-2	29
Adeno E	40.5

^aThe table contains a summary of the MFI thresholds for positive sample identification, determined after removing outliers iteratively by using the Grubb's outlier test.

Samples identified as positive via viral culture and/or immunofluorescence, including 56 RSV samples, 35 influenza virus A samples, 12 influenza virus B samples, 46 parainfluenza virus samples, and 30 adenovirus samples, were then analyzed randomly using 96-well plates, and identification performed by using multiplexed RT-PCR was compared to the viral culture and/or immunofluorescence results. A summary of this clinical study is provided in Table 4. For each respiratory virus, the table shows the number of samples identified as positive by viral culture and/or immunofluorescence, the number of samples confirmed as positive by multiplexed RT-PCR, the number of samples for which the multiplexed RT-PCR result was positive but was in disagreement with viral culture and/or immunofluorescence result, and the number of samples identified as negative by multiplexed RT-PCR. The five samples (one RSV and four parainfluenza virus samples) for which the positive diagnoses made by viral culture and by multiplexed RT-PCR were in disagreement were cultured a second time and for all 5 samples, the identification made by multiplexed RT-PCR was confirmed upon reculture. All of the samples identified positive by viral culture and/or immunofluorescence but negative by multiplexed RT-PCR (10 influenza virus A, 1 influenza virus B, 8 RSV, and 4 adenovirus samples) were sent to VRDL for third-party confirmatory q-PCR analysis. The separate singleplex semiquantitative assays run on these samples confirmed the negative multiplexed RT-PCR results for two influenza virus samples out of ten, five RSV samples out of eight, and three adenovirus samples out of four. Overall, when folding the third-party confirmatory results into the study, the multiplexed RT-PCR assay correctly identified 97% of the adenovirus, 95% of the RSV, 92% of the influenza virus B, and 77% of the influenza virus A samples without any extraction of the clinical samples (data summarized in the last column of Table 4). Compared to the combination of immunofluorescence and/or viral culture, the use of the multiplexed RT-PCR assay reduced the rate of false positives by up to 10% for adenovirus and RSV.

TABLE 4.

Summary table of the clinical study performed with the multiplexed RT-PCR respiratory panel^a

Clinical sample analysis	No. of samples:				% Reconciled multiplexed RT-PCR identifications
	Identified by viral culture and/or immunofluorescence	Confirmed by multiplexed RT-PCR	With another attribution by multiplexed RT-PCR	Negative by multiplexed RT-PCR
Negative for respiratory virus	828	791	37 (31 confirmed at VRDL)*	NA	99.3
Influenza virus A	35	25		10 (2 confirmed at VRDL)*	77
Influenza virus B	12	11		1	92
RSV	56	47	1 (confirmed after second culture)	8 (5 confirmed at VRDL)*	95
Parainfluenza virus	46	7b	4 (confirmed after second culture)	35	24
Adenovirus	30	26		4 (3 confirmed at VRDL)*	97

^aA comparison of the performance of the multiplexed assay against the results initially obtained using viral culture and/or immunofluorescence is presented for both negative and positive samples. *, all samples for which there was disagreement between the results obtained using shell vial culture and/or immunofluorescence and multiplexed RT-PCR were sent to the VRDL at the State of California Health and Human Services Agency for a third-party confirmatory q-PCR analysis. Samples were cultured a second time using standard shell vial procedure. NA, not applicable.

^bTwo parainfluenza type 1 and five parainfluenza type 3 viruses.

In order to investigate the poor performance of multiplexed RT-PCR compared to viral culture for parainfluenza virus (only 24% of correct identification on unextracted samples), we performed an extraction experiment on a subset of clinical samples diagnosed positive for parainfluenza virus by viral culture. These samples were extracted by using a magnetic bead-based viral RNA isolation kit, and 5 μl of the purified and concentrated RNA was tested by using the multiplexed RT-PCR protocol. As a control, two samples identified as parainfluenza virus type 1 and two samples identified as parainfluenza virus 3 by multiplexed RT-PCR before RNA extraction were extracted and reanalyzed in similar conditions, confirming the initial results. Three randomly selected samples initially identified as parainfluenza virus by viral culture and immunofluorescence but as negative by multiplexed RT-PCR were then extracted and reanalyzed. All three samples were identified as parainfluenza virus type 1 or 3 upon extraction, suggesting that an extraction step will be required in order to increase the sensitivity of the parainfluenza virus assay.

DISCUSSION

Although immunofluorescence and/or viral culture had initially identified 828 clinical samples as negative, 791 were confirmed negative by multiplexed RT-PCR (95.5%), while 37 were identified positive for a respiratory virus. Confirmatory q-PCR assays performed at the VRDL invalidated the positive diagnostic for 6 samples but validated it for 31 samples. Of these 31 positive samples missed when using standard detection techniques, 24 were RSV positive, 4 were influenza virus A positive, 2 were adenovirus positive, and one was parainfluenza virus positive. These data point out that most of the false negatives (77%) generated by the immunofluorescence and/or viral culture detection techniques are missed RSV samples. RT-PCR assays enabled improved detection of RSV, which could be particularly important for pediatrics departments since RSV is the most common cause of bronchiolitis and pneumonia among infants and children under 1 year of age (6).

For the analysis of the samples initially identified as positive using a combination of immunofluorescence and/or viral culture, all five samples for which there was a disagreement on the positive identification were confirmed in favor of the multiplexed RT-PCR result by a second culture. In addition, 23 samples initially identified as positive by viral culture and/or immunofluorescence were identified as negative by multiplexed RT-PCR (10 influenza virus A, 1 influenza virus B, 8 RSV, and 4 adenovirus samples). Confirmatory q-PCR analysis performed at VRDL on these samples confirmed the negative multiplexed RT-PCR results for two influenza virus, five RSV, and three adenovirus samples. The detail of the critical PCR threshold (C_T) values obtained with q-PCR for the samples that were missed using RT-PCR (one influenza virus B, one adenovirus, and eight influenza virus A samples) showed that some of these samples had fairly high C_T values after extraction, which is indicative of low levels of viral RNA in the initial sample (a C_T of 33.7 for the influenza virus B sample and a C_T above 31 for four of the missed influenza virus A samples). In addition, most of the missed samples were influenza virus A (8 of 10 missed). This can most probably be attributed to the rapid mutation rate of the influenza virus (the targeted genes for influenza virus A in our panel are the matrix protein gene M1 and the nonstructural protein gene NS) and stresses the necessity of constantly updating viral signatures to adapt the assay to the genetic evolution of the targeted organisms.

The weakness of this particular multiplexed assay is its low sensitivity to parainfluenza virus (only 7 samples [2 type 1 and 5 type 3] of 42 were detected). The LOD data pointed out that the sensitivity to parainfluenza virus type 1 was significantly weaker than the sensitivity to parainfluenza virus type 3. In order to investigate whether the signature design was the cause of the low detection levels observed for parainfluenza, parainfluenza virus type 1 and 3 clinical samples were extracted and analyzed with the multiplexed assay. Positive identification was obtained in all cases, confirming that the signatures amplify the target RNA. We therefore attribute the weakness of the parainfluenza virus assay to a lack of available free-floating RNA in unextracted samples. The viral RNA of paramyxoviruses has been reported to be encapsidated with nucleoproteins to form a very stable helical nucleocapsid (17), and a specifically strong and stable encapsidation of the parainfluenza virus RNA may be responsible for the low sensitivity of this particular assay. The parainfluenza virus type 1 signature will therefore have to be removed from this panel if no extraction is performed. Alternatively, an additional extraction step could be included in the protocol when the detection of particularly sturdy viruses is desired.

In addition to the main advantages such as flexibility, sensitivity, specificity, and ease of use, multiplexed RT-PCR also provides the ability to detect coinfections. During our clinical study, the multiplexed RT-PCR assay detected influenza virus A-adenovirus coinfections in three samples. Although not initially detected at the UCDMC, these three cases of coinfection were confirmed by a second culture. Despite the fact that only three samples showed coinfection, these results stress the unique ability of multiplexed assays to rapidly and concomitantly detect a broad range of pathogens. In addition, the high multiplexing abilities of the bead-based assays provide an effective way to reduce the cost of PCR reagents, which constitutes the main drawback of molecular diagnostics.

We have demonstrated the ability of the multiplexed respiratory panel to differentiate influenza virus from pathogens that cause influenza-like illnesses in a study including over 820 negative and over 150 positive clinical samples. The current 16-plex RT-PCR panel enables simultaneous detection of influenza virus A, influenza virus B, parainfluenza virus (types 1 and 3), RSV, and adenovirus (groups B, C, and E) in clinical samples. In order to broaden our study and increase the number of positive clinical samples, the respiratory panel is being deployed in other laboratories including the State of California Health and Human Services Agency and the Naval Health Research Center for further testing. Assay development efforts are under way to expand the capabilities of this assay by including signatures that can differentiate seasonal influenza (e.g., A/H1 and A/H3) from A/H5N1 or other potential pandemic strains. We are also in the process of developing an instrument to automate sample analysis, which will alleviate the need for skilled technicians to run the assays. This system is able to process samples, perform multiplexed real-time RT-PCR with the respiratory panel, analyze data, and report results in less than 3 h. The combination of assay development and automation should ultimately allow the implementation of the assay to perform point-of-care diagnostics, as well as disease surveillance.