Diagnostic Evaluation of Multiplexed Reverse Transcription-PCR Microsphere Array Assay for Detection of Foot-and-Mouth and Look-Alike Disease Viruses

Abstract

A high-throughput multiplexed assay was developed for the differential laboratory detection of foot-and-mouth disease virus (FMDV) from viruses that cause clinically similar diseases of livestock. This assay simultaneously screens for five RNA and two DNA viruses by using multiplexed reverse transcription-PCR (mRT-PCR) amplification coupled with a microsphere hybridization array and flow-cytometric detection. Two of the 17 primer-probe sets included in this multiplex assay were adopted from previously characterized real-time RT-PCR (rRT-PCR) assays for FMDV. The diagnostic accuracy of the mRT-PCR assay was evaluated using 287 field samples, including 247 samples (213 true-positive samples and 35 true-negative samples) from suspected cases of foot-and-mouth disease collected from 65 countries between 1965 and 2006 and 39 true-negative samples collected from healthy animals. The mRT-PCR assay results were compared to those of two singleplex rRT-PCR assays, using virus isolation with antigen enzyme-linked immunosorbent assays as the reference method. The diagnostic sensitivity of the mRT-PCR assay for FMDV was 93.9% (95% confidence interval [CI], 89.8 to 96.4%), and the sensitivity was 98.1% (95% CI, 95.3 to 99.3%) for the two singleplex rRT-PCR assays used in combination. In addition, the assay could reliably differentiate between FMDV and other vesicular viruses, such as swine vesicular disease virus and vesicular exanthema of swine virus. Interestingly, the mRT-PCR detected parapoxvirus (n = 2) and bovine viral diarrhea virus (n = 2) in clinical samples, demonstrating the screening potential of this mRT-PCR assay to identify viruses in FMDV-negative material not previously recognized by using focused single-target rRT-PCR assays.

Foot-and-mouth disease (FMD) is a highly infectious and contagious vesicular disease affecting both domestic and wild ruminants and swine; it is caused by a single-stranded positive-sense RNA virus that has seven distinct serotypes (A, Asia 1, C, O, SAT 1, SAT 2, and SAT 3) (17). FMD is endemic in many countries throughout the world, with serotype O having the highest prevalence, followed by serotype A (9). The early detection of the virus is critical to minimizing disease spread and the significant economic implications (12) resulting from the introduction of FMD into a country previously free of the disease. The diagnosis of FMD can be confounded by diseases with similar clinical signs (look-alike diseases) and by species for which the presentation of the disease is mild or indistinct (16). For the laboratory identification of FMD virus (FMDV), the Office International des Epizooties recommends virus isolation (VI), antigen enzyme-linked immunosorbent assay (Ag-ELISA), and reverse transcription-PCR (RT-PCR) with detection by agarose gel electrophoresis or in real time using TaqMan fluorogenic probes (15).

Real-time PCR is widely used by diagnostic laboratories to complement or as a replacement for more traditional detection methods. Two independent real-time RT-PCR (rRT-PCR) assays for FMD laboratory diagnosis target the ribosomal entry site of the 5′-untranslated region (5′UTR) (30) and the viral RNA polymerase gene (3D) (4) on the highly variable FMDV genome. The 5′UTR and 3D rRT-PCR assays initially were compared to each other prior to their implementation in Australia (3). A subsequent in-depth comparative evaluation was conducted to further evaluate the effectiveness of these assays, which demonstrated a higher diagnostic sensitivity of the rRT-PCR assays compared to that of VI and/or Ag-ELISA, particularly when both assays were used in combination (14). Both assays are used routinely in combination at the Food and Agriculture Organization of the United Nations, World Reference Laboratory (FAO WRL), for FMD. rRT-PCR assays also have been reported for the detection of other viruses that cause vesicular disease of livestock, including swine vesicular disease (SVD) (29), vesicular stomatitis (VS) (11, 28), and vesicular exanthema of swine (VES) (31), or symptomatic look-alike diseases, including bluetongue (13, 27, 33), bovine viral diarrhea (1, 2, 20, 22, 39), malignant catarrhal fever (37), and parapox (26).

Fluorescent probes for rRT-PCR detection have broad emission spectra that limit the multiplexing capacity to the four or five discrete optical channels typically present in most commercial real-time PCR instruments. Therefore, simultaneous testing for FMDV and look-alike disease viruses by rRT-PCR would require many assays to be run in parallel, thereby increasing the demand on instrumentation and reagents, which escalates costs. A single multiplexed screening test that simultaneously detects and differentiates FMDV from look-alike disease viruses is desirable. Such a test could facilitate the rapid and cost-effective screening of suspect FMD field samples for laboratory differential detection, targeted FMD surveillance, or embedded foreign animal disease surveillance while conducting routine testing for diseases that are endemic to those areas.

Luminex xMAP technology is a multiplexed high-throughput detection system (38) with many applications for nucleic acid detection (6). The Luminex array offers up to 100 independent channels and uses microspheres (5.6 μm in diameter) embedded with various ratios of two fluorescent dyes. User-defined surface modifications can include the addition of oligonucleotides, antibodies, peptides, or other macromolecules. Typically, a mixed suspension of functionalized microspheres is mixed with the sample to bind analytes, which then are labeled with a fluorescent reporter and analyzed using a specialized flow cytometer. For each microsphere channel, the signal resulting from the bound fluorescent reporter is measured and reported as the median fluorescence intensity (MFI), which can be compared to a cutoff value to provide end point detection with qualitative results. Recent nucleic acid applications of the Luminex array include the detection and differentiation of classical swine fever virus from other pestiviruses (5) and the typing of human respiratory viruses (18, 19, 21, 23), human papillomavirus (10, 32), and human influenza A virus (40).

This report describes a novel multiplexed RT-PCR (mRT-PCR) microsphere array assay for the differential detection of FMDV from look-alike disease viruses. The development, optimization, and analytical evaluation of this multiplex assay will be reported separately. The primary purpose of this study was to evaluate the diagnostic performance of the mRT-PCR assay for the detection of FMDV using a panel of suspect field samples. The diagnostic sensitivities of the 3D and 5′UTR FMDV assays in mRT-PCR format were compared to that of rRT-PCR using VI combined with Ag-ELISA as the reference method. The detection of look-alike disease viruses in suspect FMD field samples also is reported.

MATERIALS AND METHODS

Samples.

The panel comprised epithelia (213 true-positive and 35 true-negative samples) from suspect field cases of FMD submitted from 65 countries to the FAO WRL for FMD between 1965 and 2006 and included representatives of all seven serotypes of FMDV. In addition, 39 true-negative tongue epithelial samples were collected from healthy cattle at a United Kingdom abattoir. All sample testing and reference measurements were conducted at the FAO WRL for FMD. Epithelia were ground and suspended to generate an ∼10% (wt/vol) suspension in phosphate buffer (0.04 M, pH 7.6). The epithelial suspensions (ES) were centrifuged, and the supernatant was collected and then stored at −80°C. The FMDV serotype was determined by conducting an Ag-ELISA (8) on the original ES or after viral propagation in cell culture. Samples of look-alike viruses from the FAO WRL for FMD collection included SVD virus (SVDV), VES virus (VESV), San Miguel sea lion virus, caliciviruses isolated from a variety of species (including cetacean, bovine, feline, reptilian, and skunk species), and VS virus (VSV).

Nucleic acid extraction.

Total nucleic acid was extracted from each ES by an automated procedure using a MagNA Pure LC (Roche, United Kingdom) as previously described (14, 35). Extracted samples (40 μl) were aliquoted (three samples of 13 μl each), stored at −80°C, and thawed once just before use.

rRT-PCR.

Previously reported protocols for the individual 3D (4) and 5′UTR (34) rRT-PCR assays were modified for use in this study. Briefly, 25-μl reaction mixes (SuperScript III platinum one-step quantitative RT-PCR system [Invitrogen]) containing 20 pmol of each primer, 7.5 pmol of dual-labeled TaqMan probe, and 5 μl total nucleic acid were prepared in an optical reaction plate (Stratagene, Amsterdam, The Netherlands). For both targets, RT-PCR amplification was performed in an Mx4000 multiplex quantitative PCR system (Stratagene) as described previously (34).

mRT-PCR assay design.

A schematic depiction of the mRT-PCR assay is shown in Fig. 1. The RT-PCR uses 18 biotinylated forward and unmodified reverse primer sets (17 for detection, 1 as a control). The sequences that comprise the multiplex assay are shown in Table 1. The multiplex assay was designed to detect and differentiate FMDV from SVDV, VESV, bovine viral diarrhea virus (BVDV), bluetongue virus (BTV), parapoxviruses (PPOX; orf virus, pseudocowpox virus, and bovine papular stomatitis virus), and bovine herpesvirus type 1 (BHV-1). Primer-probe sequences for FMDV (4, 30) and BVDV (7, 22) were based on the work of others and adapted to the current multiplex format. All other sequences were designed at the Lawrence Livermore National Laboratory using an approach that was described previously (36). Additional computational analyses were performed to ensure the specificity and reliability of all available data, including a BLAST-based comparison of each primer-probe set as a triplet against all sequences in GenBank to identify the targets that are predicted to produce a PCR or TaqMan reaction at 57°C for primer annealing and 67°C for probe annealing, where temperatures are derived from Primer 3 oligonucleotide melting point calculations. Optimal candidate primer-probe sets were forwarded to the bench-screening phase for further down-selection. Amplicon sizes ranged from 95 to 349 bp.

FIG. 1.

Schematic of the mRT-PCR assay. In the presence of target nucleic acid, the biotinylated forward primer is extended during the PCR. The PCR product is hybridized to the microsphere array, at which point the extended forward primer binds to the complementary probe-labeled microsphere. The complex is labeled with a fluorescent reporter (SAPE) and then analyzed using a Bio-Plex flow cytometer. The fluorescence (570 nm) of bound reporter molecules is measured, and the MFI is calculated for each microsphere class and then compared to a cutoff to indicate the presence or absence of a target nucleic acid sequence in a sample.

TABLE 1.

Primer and probe sequences of the mRT-PCR assay^a

Assay name	Primer sequence (5′→3′)	Probe sequence (5′→3′)
BHV-1	Forward, GT-GCCAGCCGCGT-AAAAG	TCCTGGTTCCAGAGCGCTAACATGGAG
	Reverse, GACGACTCCGGGCTCTTTT
BHV-2	Forward, TGAGGCCT-ATGTATGGGCAGT-T	AAATAACACGGTGTGCACTTAAATAAGATTCGCG
	Reverse, GCGCGCCAAACATAAGTAAA
BTV-1	Forward, GCACCCT-ATATGTTT-CCAGACCA	CTAACTCGTGGGCCAATCATCATCTTCTGT
	Reverse, CAGCTAACTCTTCAGCCACACG
BTV-2	Forward, AGAATT-CAGGAT-GGGCAGGA	CCATCACACCATTATACTGTACCCGCGTAGC
	Reverse, GCACAATTCCCATCCCCTTA
BVDV	Forward, GGTAGTCGT-CAGTGGTT-CGAC	CCTCGTCCACGTGGCATCTCGAG
	Reverse, CATGTGCCATGTACAGCAGAGAT
FMDV 3D	Forward, ACTGGGT-TTTACAAACCT-GTGA	GTCCCACGGCGTGCAAAGGA
	Reverse, GCGAGTCCTGCCACGGA
FMDV 5′UTR	Forward, CACYTYAAGRT-GACAYTGRTACT-GGTAC	CCTCGGGGTACCTGAAGGGCATCC
	Reverse, CAGATYCCRAGTGWCICITGTTA
PPOX-1	Forward, GCAGAT-GCGCTCCT-GGTT	CCGACTCCGACGTGGAGAACGTG
	Reverse, GCACCTCTGCTGCTGCAA
PPOX-2	Forward, GATGGCCGT-GCAGCT-CTT	TGTACGGGCTCATGGGCTTCCG
	Reverse, CGTACAAGATCACGGCCAACT
PPOX-3	Forward, GCAGCAGT-GCACCACGT-AGT	GACTTCGAGGCGGACAACAAGCG
	Reverse, CGCTGAACCCGTACATCCT
SVDV-1	Forward, CAGGAT-AATTTCTT-CCAAGGGC	TGCATTGTGTCTGATGGTACAACTTGTGACG
	Reverse, ACGTGAACATTTCGAGCTTCC
SVDV-2	Forward, GACTTGT-TGTGGCT-GGAGGA	TGACCGTAATGAGGTCATCGTGATTTCTCAC
	Reverse, CAGCGCCATGGTGAGGTAG
SVDV-3	Forward, GACAAAGT-GGCCAAGGGAAA	CTGGCGTCATAGCCTGAATAGTCAAACGCTA
	Reverse, CACGTAAACCACACTGGGCT
VESV-1	Forward, GCCTT-CTCCCTT-CCCAAAA	CATCATCGTTGATAACCTTAGATGTGCAATTTGG
	Reverse, TGAAGGAATGGTTCCGTCAGT
VESV-2	Forward, GGGAAT-GAGGTGTGCAT-CATT	AAATTGGCATAATCAACCTTGTCAGATGAGTCG
	Reverse, CACGTCTTGATGTTGGCTTGAC
VESV-3	Forward, GGTCGCT-CTCACTGATGAT-GAGTA	GCTCGGTGCCTGAGTTGGAGGAAG
	Reverse, GGTGTTATCAGCACCCATTGC
VESV-4	Forward, ACCACCT-CTGGAAACATCT-ATGG	CGGGACGGGCATTTGTCACCA
	Reverse, TTTGTGCACGTGTCACGAAT
FC	NA	CAAAGT-GGGAGACGTCGT-TG
IC	NA	CAAAGTGGGAGACGTCGTTG-Cy3
NC	NA	CAAAGTGGGAGACGTCGTTG

^aY = pyrimidines (C/T), R = purines (A/G), W = weak two-bonds (A/T), I = inosine (universal base). , an internal biotinylated d(T). All forward primers also include biotinylation at the 5′ terminus. All probes contain an amine attached to the 5′ terminus with C-6 and internal spacer 18. NA, not applicable.

mRT-PCR primers and probes.

All oligonucleotides used for mRT-PCR were synthesized by Integrated DNA Technologies (Coralville, IA) and were purified by high-performance liquid chromatography. Forward primers were functionalized with 5′-terminal and internal biotin moieties. Reverse primers were unmodified. The probes were modified by 5′-amino C-6 (phosphoramidite) with an 18-atom hexaethyleneglycol spacer. The lyophilized probe was dissolved in 2-(N-morphilino)ethanesulfonic acid (MES) to yield a stock concentration of 1 mM. Lyophilized forward and reverse primers were dissolved in TE buffer (100 μl; 1 mM Tris, 10 mM EDTA, pH 7.4) to yield a stock concentration of 1 mM. Working dilutions were prepared from the stock solutions as required.

Coupling of probe oligonucleotides to microspheres.

xMAP multianalyte COOH microspheres (Luminex Corp., Austin, TX) were covalently coupled to probe oligonucleotides using carbodiimide activation based on the manufacturer's protocol. Briefly, stock microspheres (1 ml; 1.25 × 10⁷ microspheres) were vortexed for 30 s, sonicated for 60 s, and centrifuged at 8,000 × g for 5 min, and then the supernatant was removed. The microspheres then were resuspended in MES (50 μl, 0.1 M, pH 4.5), vortexed, and sonicated. Probe (10 μl; 50 μM in MES) was added, and the mixture was vortexed. An aqueous solution of N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC; 5 μl, 10 mg/ml) was added, vortexed, and then gently agitated for 30 min in the dark. A second aliquot of EDC (5 μl, 10 mg/ml) was added, vortexed, and then gently agitated for 30 min in the dark. Tween 20 (1 ml, 0.02% [vol/vol]) was added, vortexed, and centrifuged, and the supernatant was removed, after which the washing was repeated using sodium dodecyl sulfate (1 ml, 0.1% [mass/vol]) and then TE buffer. The probe-conjugated microspheres were resuspended in TE buffer (250 μl), vortexed, and then stored at 4°C in the dark.

Microsphere mixture.

A 21-plex microsphere suspension was prepared by combining individual stock suspensions (9 μl/class) with Tris-NaCl buffer (3 ml; 0.1 M Tris, 0.2 M NaCl, 0.05% [vol/vol] Triton X-100, pH 8.0; Teknova). The mixture was vortexed and then enumerated using the Bio-Plex workstation (Bio-Rad, CA), targeting approximately 150 microsphere counts per class in 40 s. If required, additional microspheres from the individual stocks were added to ensure that the concentrations of all classes were approximately equal.

mRT-PCR amplification.

Each field sample was analyzed in duplicate by the multiplex assay. Amplification was performed using a one-step RT-PCR kit (SuperScript III one-step RT-PCR system with platinum Taq DNA polymerase; Invitrogen). The reaction volume of 25 μl was comprised of nuclease-free water (0.95 μl), primer mix (3.6 μl), SuperScript III 2× reaction mix (12.5 μl), MgSO₄ (0.95 μl, 50 mM; Invitrogen), SuperScript III RT/platinum Taq mix (1 μl), internal control armored RNA (1 μl; ∼100 copies), and template (5 μl). The internal control armored RNA was lysed by being heated to 70°C for 4 min prior to being added to the mix. The final concentration of each primer and MgSO₄ was 0.4 μM and 3.5 mM, respectively. The mRT-PCR thermal cycling protocol was 55°C for 30 min, 95°C for 2 min, and then 35 cycles of 95°C for 15 s, 60°C for 30 s, and 72°C for 15 s, followed by 72°C for 2 min with a final hold at 4°C.

Microsphere array hybridization.

A wash-assay format was adopted to reduce the variability of the response caused by the nonspecific hybridization of PCR products and fluorescent label. In a 96-well plate, RT-PCR product (1 μl) was added to a mixed suspension of probe-conjugated microspheres (22 μl) and then placed in a thermal cycler and subjected to 95°C for 2 min, 55°C for 5 min, and then a 4°C hold. Tris-NaCl (100 μl) was added, and the suspension was transferred to a 96-well vacuum filter plate (MABVN 1250 multiscreen filter plate; Millipore). The suspension was vacuum aspirated and washed twice with Tris-NaCl buffer (100 μl for each wash). Stock streptavidin phycoerythrin (SAPE; 1 mg/ml; Caltag Laboratories) was diluted with Tris-NaCl to a working concentration of 3 μg/ml. SAPE (60 μl, 3 μg/ml) was added, and the suspension was incubated in the dark for 5 min. The suspension was vacuum aspirated, washed once with Tris-NaCl (100 μl), resuspended in Tris-NaCl (100 μl), and then transferred to a 96-well round-bottom plate for fluorescent detection. Nucleic acid extraction, PCR assembly, and PCR amplification were conducted in separate rooms. Likewise, hybridization and Bio-Plex detection were performed together in a separate room to minimize the likelihood of PCR contamination by amplicons.

mRT-PCR detection.

Fluorescence detection of the processed microsphere suspension array was achieved using a Bio-Plex workstation set to count a minimum of 100 events per microsphere class in a 50-μl Bio-Plex sample volume. The Bio-Plex workstation is a specialized dual-laser flow cytometer integrated with an XY microplate platform configured to analyze Luminex xMAP microspheres in a 96-well plate format. The reporter photomultiplier tube voltage was calibrated on the low setting with background subtraction enabled. The Bio-Plex workstation was validated and calibrated per the manufacturer's instructions. The resolution of the MFI was 0.5 U for all channels.

mRT-PCR assay controls.

The multiplex assay incorporates four control channels integral to each reaction that are used to verify the assay's integrity. The negative control (NC) is a microsphere conjugated to a Thermotoga maratima-derived oligonucleotide sequence (MT-7) that serves as a nonspecific binding control in the mRT-PCR assay, and its response should remain consistently low (MFI ≤ 80). The fluorescence control (FC), biotinylated MT-7, confirms that fluorescent labeling with SAPE occurred, and it should exhibit a high response (MFI > 1,000). The instrument control (IC) comprises a Cy3-labeled MT-7 conjugate. As Cy3 and SAPE have similar fluorescence excitation and emission wavelengths, the IC confirms the proper function of the reporter optics within the Bio-Plex flow cytometer (MFI > 500). Armored RNA served as an end-to-end amplification control (AC) to reduce the probability of a false-negative result and is utilized at low concentrations (100 copies/reaction) to generate a low-level response (MFI ≥ 20) that minimizes its competition with detection channels. The armored RNA (XenoRNA-01; Ambion, Austin, TX) is a proprietary 1,070-nucleotide RNA transcript consisting of unique nucleotide sequences that possess no significant homology to the current annotated sequences in commonly used sequence databases, including NCBI, Affymetrix, and Rosetta. Primers and probe-labeled microspheres for the AC are included in the multiplex primer mix and microsphere suspension, respectively.

mRT-PCR assay integrity.

Minimum bead count quotas and control channel responses were used to verify the integrity of the detection channel responses, which minimizes the likelihood of false-positive and false-negative results caused by operator error, instrument malfunction, nonspecific hybridization, or PCR inhibition. The responses of all detection channels are checked against preestablished cutoffs, after which the results are defined as mRT-PCR positive, negative, or inconclusive. For each sample, the MFI of each control bead class was checked against a cutoff value. A given result was considered invalid if the MFI value of the IC, NC, or FC control was out of range or if both the response of the AC was <20 and no detection channels exceeded the cutoff. Results were considered valid when the AC was <20 and any detection channel exceeded the cutoff. The AC response can be diminished by a strong positive response on a detection channel caused by competition in the RT-PCR. If the responses of all control channels are acceptable, the numbers of beads counted for control and detection channels are checked (≥40 beads per channel). If a low bead count for any control channel occurred, all results for that sample were considered invalid. A low bead count on a given detection channel was considered an invalid result for the channel in question.

Data analysis.

Raw data exported from the Bio-Plex instrument were imported into MATLAB (MathWorks) and then analyzed using Microsoft Excel. If at least one result from a duplicate sample analysis exceeded the cutoff, the sample was assigned as mRT-PCR positive. Receiver operating characteristic (ROC) plots (24) were generated using a custom MATLAB program. Published guidelines (25) were followed for the calculations of diagnostic test accuracy and for statistical methods to quantify uncertainty.

Cutoff values.

The mRT-PCR responses for this assay typically are non-Gaussian; therefore, nonparametric methods were used to determine cutoffs. Each channel in the multiplex has a distinct distribution of responses to true-negative samples; therefore, each is assigned its own cutoff value. For each detection channel of the mRT-PCR assay, the responses of true-negative samples were ranked according to magnitude, and then cutoff values were identified as the response (MFI value) that gave a false-positive rate closest to 5% (without exceeding 5%), corresponding to a diagnostic specificity of at least 95% (3D MFI, ≥6.5; 5′UTR MFI, ≥5.5). Other mRT-PCR detection channels for the FMD look-alike disease viruses had similar cutoffs. For rRT-PCR, the cutoff was a cycle threshold (C_T) of ≤32 for both 3D and 5′UTR assays (34).

RESULTS

Clinical sample validation.

ROC plots were constructed to analyze the diagnostic performance of the 3D and 5′UTR assays in rRT-PCR and mRT-PCR formats (Fig. 2). For the purpose of this evaluation, the presence of FMDV was designated using VI and/or Ag-ELISA as the reference method(s), defining samples as true negative or true positive, with the caveat that rRT-PCR is known to detect FMDV in some samples considered negative by these methods (35). Each plot shows the true-positive fraction (sensitivity) versus the false-positive fraction (1 − specificity) over the entire range of cutoff values. These ROC plots indicate that the 3D and 5′UTR assays in the mRT-PCR format lost some of their ability to distinguish between true-negative and true-positive samples compared to the ability of rRT-PCR to do so. The associated areas under each ROC plot also indicate the level of test performance in the absence of a cutoff value. Transfer to the multiplex format had the greatest effect upon the 5′UTR assay, for which the area under the curve was reduced to 0.773, whereas that for the rRT-PCR format was 0.942. In contrast, the effect upon the 3D assay was less apparent: the areas under the curves were 0.955 and 0.985 for the mRT-PCR and rRT-PCR formats, respectively. For the purposes of this analysis, these ROC plots for the two targets were considered independently, although in practice the results for 3D and 5′UTR assays are combined in the mRT-PCR format. The trade-off between the true-positive fraction and false-positive fraction shown by the ROC plots can be used to inform cutoff selection. The cutoff value for each detection channel was calculated from the distribution of mRT-PCR responses to true-negative samples (n = 74) using a specificity of 95%.

FIG. 2.

ROC plots for the 3D and the 5′UTR assays in singleplex rRT-PCR and mRT-PCR formats. True-positive samples (n = 213) representative of all FMDV serotypes and true-negative samples (n = 74) were analyzed. The areas under the curves are the following: 3D rRT-PCR, 0.985; 5′UTR rRT-PCR, 0.942; 3D mRT-PCR, 0.955; and 5′UTR mRT-PCR, 0.773.

Table 2 summarizes the performance metrics of the 3D and 5′UTR assays when used independently or in combination for both formats. The results are presented according to the serotype and then are summarized for all serotypes. In some cases, the individual rRT-PCR response for true-positive samples (3D, n = 5; 5′UTR, n = 9) and true-negative samples (3D, n = 5; 5′UTR, n = 2) yielded C_T values that were beyond the cutoff. In practice, these samples would be considered weak rRT-PCR positives and retested.

TABLE 2.

Performance metrics for the 3D and 5′UTR FMDV assays in rRT-PCR and mRT-PCR formats using independent or combined results

Performance metrica	rRT-PCR			mRT-PCR
	3D	5′UTR	Combined	3D	5′UTR	Combined
% Sensitivity (fraction)	93.5 (43/46)	87.0 (40/46)	97.8 (45/46)	80.4 (37/46)	76.1 (35/46)	93.5 (43/46)
% Sensitivity by serotype (fraction)
Asia 1	100 (10/10)	100 (10/10)	100 (10/10)	100 (10/10)	100 (10/10)	100 (10/10)
C	95.0 (19/20)	95.0 (19/20)	95.0 (19/20)	85.0 (17/20)	55.0 (11/20)	90.0 (18/20)
O	96.7 (87/90)	86.7 (78/90)	97.8 (88/90)	91.1 (82/90)	70.0 (63/90)	92.2 (83/90)
SAT 1	100 (18/18)	88.9 (16/18)	100 (18/18)	100 (18/18)	16.7 (3/18)	100 (18/18)
SAT 2	100 (21/21)	71.4 (15/21)	100 (21/21)	95.2 (20/21)	9.5 (2/21)	95.2 (20/21)
SAT 3	100 (8/8)	100 (8/8)	100 (8/8)	100 (8/8)	12.5 (1/8)	100 (8/8)
All	96.7 (206/213)	87.3 (186/213)	98.1 (209/213)	90.1 (192/213)	58.7 (125/213)	93.9 (200/213)
95% CI	93.4-98.4	82.2-91.1	95.3-99.3	85.4-93.5	52-65.1	89.8-96.4
Specificity (%)	94.6	95.9	93.2	93.2	94.6	91.9
95% CI	86.9-97.9	88.7-98.6	85.1-97.1	85.1-97.1	86.9-97.9	83.4-96.2
Sample disease prevalence (%)	74.2	74.2	74.2	74.2	74.2	74.2
PPV (%)	98.1	98.4	97.7	97.5	96.9	97.1
NPV(%)	90.9	72.4	94.5	76.7	44.3	84.0
Efficiency (%)	96.2	89.5	96.9	90.9	67.9	93.4

^aSensitivity (fraction) means the fractions of the 213 true-positive epithelial tissue samples (subjected to both mRT-PCR and rRT-PCR assays) that are classified as FMDV positive by the indicated assay. CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value.

Table 3 shows a three-way comparison of mRT-PCR and rRT-PCR to the reference methods using the combined 3D and 5′UTR assay results. The agreement between combined rRT-PCR and combined mRT-PCR results for true-positive samples (n = 213) was 95.8% (204/213), including 200/213 samples for which both combined assays gave a positive result and 4/213 samples for which both assays gave a negative result. Two samples (SYR 6/2002 and LAO 16/2003) classified as FMDV negative by VI and Ag-ELISA tested positive by both RT-PCR formats. Similarly, three samples (SYR 7/2002, TUR 17/2002, and BHU 5/2004) that were negative by VI and Ag-ELISA tested positive by rRT-PCR but were negative by mRT-PCR due to its higher limit of detection (LOD). These findings are consistent with those of earlier studies (14, 35), in which the higher analytical sensitivity of rRT-PCR enabled the detection of FMDV in samples designated negative by VI and Ag-ELISA. The higher LOD of the mRT-PCR caused it to miss 9/209 positive samples detected by rRT-PCR. Four mRT-PCR false positives had MFI responses that were close to the cutoff, which was defined using a specificity of 95%. Four mRT-PCR and rRT-PCR false negatives were samples of FMDV serotypes A (NIG 12/74), C (PHI 2/89), and O (YEM 1/2001 and a weak titrated sample of the FMDV reference isolate O₁ Manisa [TUR 8/69]). However, the rRT-PCR did generate C_T responses for two of these samples, PHI 2/89 (3D, 38.94; 5′UTR, 38.51) and O₁ Manisa (3D, 32.71), although they were weaker than the cutoff used to define positives. These false negatives almost certainly are explained by the poor quality or small amounts of virus present in these samples. In support of this conclusion was the finding that no virus was isolated in cell cultures from two of these samples (PHI 2/89 and YEM 1/2001), although Ag-ELISA was able to detect FMDV. Furthermore, the original analysis of NIG 12/74 required a second passage in cell culture to isolate FMDV, indicating the presence of a small amount of virus.

TABLE 3.

Three-way comparison for the combined results of 3D and 5′UTR assays in rRT-PCR and mRT-PCR formats

Total no. of samples	Method and result		True diagnosis
	Combined rRT-PCR	Combined mRT-PCR	Positive	Negative
202	Positive	Positive	200	2
12	Positive	Negative	9	3
4	Negative	Positive	0	4
69	Negative	Negative	4	65
Total
287			213	74

In addition to the 287 samples used to evaluate the mRT-PCR assay, 11 suspected FMDV field samples of undetermined serotype were analyzed. These samples were found to be negative by VI and Ag-ELISA but confirmed positive by combined 3D and 5′UTR rRT-PCR during previous studies (14, 35). These samples were analyzed by the mRT-PCR assay, which detected 11/11 of these samples, demonstrating that the multiplex also detects FMDV in clinical samples that most likely were rendered nonviable for VI between collection and laboratory receipt.

LOD.

The LODs of the 3D and 5′UTR assays in mRT-PCR and rRT-PCR formats were compared using serially diluted clinical samples of serotype O or SAT 2 (Fig. 3). At the defined cutoff values, the mRT-PCR LOD was higher than that for rRT-PCR by approximately 5 to 625 times for the 3D assay and 25 to 125 times for the 5′UTR assay, depending on the serotype. The higher LOD of the mRT-PCR therefore caused the loss of diagnostic performance, which was evident in the ROC plots. Further assay optimization hopefully could improve the limits of detection of the 3D and 5′UTR assays in the mRT-PCR format.

FIG. 3.

Comparison of the 3D and 5′UTR assay responses in mRT-PCR and singleplex rRT-PCR formats by consecutive fivefold serial dilutions of three different FMDV true-positive epithelial suspension samples (undiluted to more dilute from right to left). Detection by mRT-PCR was less sensitive than that by singleplex rRT-PCR (approximately 5 to 625 times and 25 to 125 times for 3D and 5′UTR signatures, respectively). Cutoff values for rRT-PCR (C_T ≤ 32) and mRT-PCR (3D MFI, ≥6.5; 5′UTR MFI, ≥5.5) are indicated by the vertical and horizontal dashed lines, respectively. The 5′UTR signature did not respond to the SAT 2 (SAU 4/2000) sample at any dilution in singleplex rRT-PCR or mRT-PCR formats and was omitted from the plot. Error bars indicate ±1 standard deviation of the mean (n = 2) response from the signature in the mRT-PCR format. For rRT-PCR, each sample was analyzed as a singlet (n = 1). Pos., positive; Neg., negative.

Differential detection.

The characterization of the diagnostic accuracy of the look-alike disease assays in the mRT-PCR assay is ongoing and will be reported separately. The mRT-PCR assay correctly identified the virus present in three SVDV field samples (ITL 4/77, HKN 1/80, and HKN 5/91), two VESV isolates (serotypes B51 and H54), five San Miguel sea lion virus isolates (serotypes 7, 9, 10, 11, and 13), and two cetacean calicivirus isolates (Tur-1 and dolphin). The mRT-PCR did not detect VESV-B1-34 and other VESV serotypes, including bovine (Bos-1 [Tillamook]), feline (A4), reptilian (rattlesnake), and skunk serotypes and also VSV (serotype NJ 15/88 CP211634 and Indiana 1 subtype Ind 2 Maipu Argentina). Feline calicivirus is in a genomic group distinct from the other VESVs tested and was not expected to be detected by this assay. mRT-PCR also identified look-alike disease viruses in four suspect FMD field samples that previously had been designated FMDV negative by VI and rRT-PCR (Fig. 4). Two samples from cattle (IRN 4/2002 and IRQ 58/2002) were mRT-PCR positive for PPOX. The PPOX-3 assay generated the strongest response; however, all three assays exceeded their respective cutoff values for both samples. Two further cattle samples (UKG 36/94 and UKG 37/94) were mRT-PCR positive for BVDV and also were confirmed to be FMDV negative by all methods. The mRT-PCR simultaneously detected the coinfection of FMDV and BVDV in an FMDV true-positive sample (HUN 2/72). The presence of BVDV, which causes a disease prevalent among cattle, did not mask the detection of FMDV.

FIG. 4.

mRT-PCR assay identification of FMDV look-alike disease viruses in suspect FMDV clinical sample submissions. The mRT-PCR assay ruled out FMDV while simultaneously ruling in FMD look-alike disease viruses. (A and B) Multiloci detection of PPOX. (C) BVDV detection in field samples from cattle (UKG 36/94 and UKG 37/94) that tested FMDV negative by virus isolation and rRT-PCR and the simultaneous detection of FMDV and BVDV (HUN 2/72). mRT-PCR cutoffs indicated by the vertical dashed lines were the following: PPOX-1, ≥7.5; PPOX-3, ≥9.5; and BVDV, ≥6.5. The horizontal dashed lines indicate the cutoff for the FMDV 3D assay (≥6.5). Pos., positive; Neg., negative.

Cross-talk between detection channels was minimal, even at the relatively high concentrations of FMDV RNA in many of the field samples tested. A matrix of correlation coefficients for all channel pairings was calculated using the results of all FMDV true-positive samples (n = 213). The correlation coefficients of the 3D and 5′UTR channels with other channels did not exceed 0.209 and 0.114, respectively.

DISCUSSION

The mRT-PCR assay was able to detect FMDV at clinically relevant concentrations. The apparent concentration of FMDV in the clinical samples was relatively high, as might be expected with vesicular epithelial tissue, which, when infected, is rich in virus. The performance of the mRT-PCR and that of the singleplex rRT-PCRs were compared using VI and/or Ag-ELISA to define FMDV true positives and true negatives. Although VI and Ag-ELISA are established methods for the detection of FMDV, previous studies (14, 35) have demonstrated that rRT-PCR has a higher diagnostic sensitivity and can detect virus in additional samples, which for the purposes of this study would be classified as true negatives. For the majority of true-positive field samples, the 3D mRT-PCR response was saturated and grouped far from the cutoff. The 5′UTR mRT-PCR signal generally was lower than that of the 3D mRT-PCR and clustered on either side of the cutoff. The higher LOD of both assays in the mRT-PCR format may be caused by low-level nonspecific interactions between primer sets that could reduce the amplification efficiency. Furthermore, the primers used in the 5′UTR assay were not originally designed with multiplexing in mind. In order to recognize a wide range of FMDV isolates, the 5′UTR primers have a high degree of degeneracy (32- and 8-fold for the forward and reverse primers, respectively, which indicates the number of oligonucleotide sequence variations, due to degenerate bases, that are present in the PCR [30]) that could increase the likelihood of nonspecific interactions with other primer sets in the multiplexed reaction mixture. While primer-probe sets incorporating degenerate nucleotides can offer broader coverage of highly variable gene segments, further work is required to refine their design for multiplexed assays. Asymmetric PCR and multivariate optimization may lead to further improvements in the LODs of the 3D and 5′UTR mRT-PCR assays. For example, Deregt et al. (5) recently showed that for an mRT-PCR Luminex array assay for the detection of classical swine fever and other pestiviruses, asymmetric PCR generated MFIs that were much higher than those obtained by symmetric PCR. Asymmetric PCR preferentially generated single-stranded DNA products through the extension of the higher-concentration biotinylated reverse primer, which in turn increased its binding efficiency to the complementary probe-labeled microsphere. A similar approach could be considered for further optimization of the current assay.

The 3D and 5′UTR assay responses had serotypic bias, a finding that agrees with earlier observations (4, 14). In the mRT-PCR format, the 3D assay was less sensitive for A and C serotypes, whereas the 5′UTR was less effective against SAT serotypes. For all serotypes collectively, the 3D assay was more sensitive than the 5′UTR assay within each format (rRT-PCR 3D and 5′UTR sensitivities, 96.7 and 87.3%, respectively; mRT-PCR 3D and 5′UTR sensitivities, 90.1 and 58.7%, respectively). In an earlier evaluation of the rRT-PCR assays (14), the diagnostic sensitivity of the 3D assay (97.7%) was found to be slightly higher than that of the 5′UTR assay (95.4%). The diagnostic sensitivity increased when the results of the 3D and 5′UTR assays were combined. This is due to the “or” nature of the combination, in which a single mRT-PCR positive result on the 3D or 5′UTR channel generates a combined mRT-PCR positive result. The diagnostic sensitivity of the combined mRT-PCR assay was 93.9%; that of the combined rRT-PCR was 98.1%. The loss of diagnostic sensitivity between rRT-PCR and mRT-PCR, due to a higher LOD, was partially offset by the inherent ability of the multiplex assay to simultaneously screen multiple loci. Conducting parallel rRT-PCR assays showed that combining results from both rRT-PCR assays increased the diagnostic sensitivity by only 1.4% above that of using the 3D rRT-PCR assay results in isolation. As the mRT-PCR is a screening assay that most likely would be used in conjunction with confirmatory tests, potential users may be more tolerant of lower specificity in order to achieve higher sensitivity.

The differential detection of FMDV from look-alike disease viruses, which included single-stranded RNA and double-stranded DNA targets, was demonstrated by testing representative isolates of SVDV, VESV, and VSV. For a diagnostic laboratory, this could produce time and cost savings compared to the time and cost of testing for each disease using singleplex rRT-PCR assays. For veterinarians, the mRT-PCR assay could increase confidence in a sample identified as FMDV negative by simultaneously screening for the presence of look-alike diseases. For networks of veterinary diagnostic laboratories, an mRT-PCR assay could facilitate embedded foreign animal disease surveillance while conducting routine testing for animal disease viruses that are endemic to the area. An interlaboratory evaluation of this multiplex assay recently was conducted in 14 U.S. National Animal Health Laboratory Network laboratories; the results suggested that the mRT-PCR technology could be operated successfully in this setting. The diagnostic performance evaluation for the look-alike disease assays in the mRT-PCR format is under way and will be reported separately.

The mRT-PCR format is compatible with the procedures and instrumentation used for rRT-PCR. The use of a single method to prepare clinical samples for mRT-PCR and rRT-PCR analysis was demonstrated in this study. The mRT-PCR requires the postprocessing of RT-PCR products to the microsphere array, which takes ∼50 min per 96-well plate using manually operated multichannel pipettes. The 96-well plate format provides convenient interchangeability between manual and automated platforms. With this reagent set, the Bio-Plex flow cytometer analyzed each well in ∼40 s, or ∼1 h per 96-well plate. Although the diagnostic sensitivity of the mRT-PCR for FMDV detection is lower than that of the singleplex rRT-PCR, it provides significantly more diagnostic information. With 17 detection channels for seven different viruses, the current prototype panel generates 1,632 individual assay results per 96-well plate. The microsphere suspension array is a versatile platform compatible with many different types of diagnostic tests, including immunological and serological assays, which could increase its utility within a veterinary diagnostic laboratory not only for outbreak response and recovery but also for routine testing for diseases that are endemic to the area. The inherent flexibility of the Luminex array also enables the composition of a given multiplex assay to be altered by simply adding or removing detection channels. Refinements to the first version of the mRT-PCR assay described herein currently are under way, including the development of two species-specific panels for more comprehensive coverage. These new panels incorporate additional assays for other FMDV look-alike diseases. The bovine-specific panel incorporates assays for FMD, malignant catarrhal fever, rinderpest, bluetongue, BHV-1 disease, bovine viral diarrhea, parapox, and VS. The porcine-specific panel includes assays for FMD, SVD, VES, VS, and porcine reproductive and respiratory syndrome.