Serum samples for the clinical performance studies were obtained from the Dengue Surveillance System sample collection under guidelines approved by the CDC institutional review board (IRB). Patients did not provide verbal or written consent for this study. Samples were de-linked from patient identifiers according to CDC's IRB.

Several DENV panels were used to evaluate the analytical performance of the RT-PCR Assay. A 5 replica panels of eight 1∶10 dilutions of quantified (GCE/mL) laboratory-adapted, heat inactivated DENV strains (DENV-1 Haw, DENV-2 NGC, DENV-3 H87, DENV-4 H241) were prepared in human serum or plasma (Bioreclamation Inc). A second panel consisted of 29 DENV-1–4 isolates obtained from serum samples from patients residing in different countries world-wide, cultured, and quantified as described previously [46]. Briefly, serum was inoculated into C6/36 (Aedes albopictus) cell culture maintained in Dulbecco's Minimal Essential Medium and incubated at 33°C for 5 days. The quantified stocks were diluted in human serum at 1∶10 dilutions down to 10² GCE/mL. Triplicate samples of each dilution were tested. Serum samples were collected under guidelines approved by the CDC IRB.

In vitro-transcribed RNA was used as the copy number control to quantify stocks of laboratory-adapted DENV strains as previously described [47]. Briefly, templates were developed by amplifying DENV-1–4 using the same corresponding primer pairs in the CDC DENV-1–4 Real Time RT-PCR Assay (Table 1) and cloned into a TOPO TA vector (Invitrogen, cat# K4600-01). Target RNA was transcribed with T7 RNA polymerase using AmpliScribe T7 Flash Transcription Kit (Epicentre, cat# ASF3257). The resulting RNA was quantified by spectrophotometry and expressed as genome copy equivalents per mL (GCE/mL).

The Assay includes a set of oligonucleotide primers and dual-labeled 5′-fluorescent (Taqman) probes for in vitro, qualitative detection of DENV-1–4. The targeted regions of DENV RNA are transcribed into complimentary DNA (cDNA) and amplified by the polymerase chain reaction (PCR). The fluorophore-labeled probes anneal to amplified DNA fragments and the fluorescent signal intensity is monitored by the ABI 7500 Fast Dx instrument during each PCR cycle. Target amplification is recorded as increase and accumulation of fluorescence over time in contrast to background signal. The assay can be run in singleplex (each DENV serotype detected in a separate reaction) or in multiplex (the four DENV serotypes are run in the same reaction) formats facilitated by the targeting of each DENV serotype with a different colored probe, i.e. each Taqman probe targets a single DENV serotype and is conjugated to a fluorophore that emits fluorescence at different excitation wavelengths: 5′-FAM DENV-1 BHQ1-3′, 5′HEX DENV-2 BHQ1-3′, 5′-TR DENV-3 BHQ2-3′, and 5′-Cy5 DENV-4 BHQ3-3′. All validations in this study were performed using the CDC DENV-1–4 Real Time RT-PCR Assay in accordance to the manufacturer's instructions (Package Insert, KK0128 available at www.cdc.gov/dengue). Viral RNA was extracted using the QIAamp DSP Viral RNA Mini kit (Qiagen cat# 61904) following manufacturer's suggested protocol. All RT-PCR reactions were performed and validated using SuperScript III Platinum One-Step Quantitative RT-PCR System without Rox (Invitrogen, cat# 11732-088). To assemble a multiplex RT-PCR reaction, 5 µL of RNA were mixed with the following reagents: 2.2 µL of nuclease-free H₂O, 12.5 µL of 2× premix, 0.5 µL of forward and reverse primers for DENV-1 and DENV-3 (final concentration 1 µM), 0.25 µL of forward and reverse primers for DENV-2 and DENV-4 (final concentration 500 nM), 0.45 µL of each Taqman probe (final concentration 180 nM), and 0.5 µL of SuperScript III RT/Platinum Taq mix to a final reaction volume of 25 µL. Individual reactions were run in either 8-tube optical strips or 96-well plates and placed in the ABI 7500 FAST Dx thermocycler (Applied Biosystems, cat# 4406985). The standard cycling method was selected and fluorescence capture set to detect emissions through the FAM, HEX, Texas Red, and CY5 channels in each well. Thermocycling parameters were as follows: reverse transcription (RT) at 50°C for 30 min, RT inactivation at 95°C for 2 min and fluorescence detection for 45 cycles of 95°C for 15 seconds and annealing at 60°C for 1 min. Amplification curves were evaluated by serotype and the threshold line placed above overt background signal, usually intersecting the initial exponential phase of the curve for each serotype individually. Amplification curves with CT values >37 render erratically and are difficult to ascertain with increasing CT values; therefore present unreliable results and were considered negative.

Several viral RNA extraction methods, manual or automated, were validated for use with this assay. The QIAamp DSP Viral RNA Mini kit (Qiagen, cat# 9001292), used manually or in the automated QIAcube and the MagNA Pure LC 2.0 automated system (Roche, cat# 05-197-686-001) were extensively validated. Additional kits and platforms were tested and found to have similar performance, including QIAamp Viral RNA mini kit (Qiagen, cat#s 52904 and 52906) manual or on QIAcube; the automated BioRobot Universal System (Qiagen, cat# 9001094); the automated BioRobot M48 System (Qiagen, cat# 9000708).

Sequencing of the DENV envelope glycoprotein (E) gene was performed as a comparative method to confirm diagnostic results obtained with the assay. The E gene (1,485 bp) from every isolate was sequenced. Briefly, E gene was amplified using BluePrint One-Step RT-PCR kit (Takara #RR755A) with serotype-specific primers following manufacturer suggested protocol. The sequences were determined by BigDye Terminator v3.1 Cycle sequencing kit (Life Technologies, cat#4337455) following manufacturer suggested protocol and bi-directional Sanger method. All amplification and sequencing primers are listed in Supporting Table S1. The coding sequences were assembled using SeqMan found in DNASTAR Lasergene 9 Core Suite (http://www.dnastar.com/t-products-lasergene.aspx). All sequence alignments were performed using the ClustalW module from MEGA5 (www.megasoftware.net). Phylogenetic analyses were performed by generating maximum likelihood trees under the GTR+I+Γ4 nucleotide substitution model available in MEGA5. GenBank accession numbers are available upon request.

A total of 86 acute serum samples were collected from febrile patients in Puerto Rico and Costa Rica (2009–2011) suspected of having dengue 0–5 days after onset of symptoms and whose average age was 14.3 years, in a study approved by the CDC IRB. Viral RNA was extracted using the Qiagen QIAamp DSP Viral RNA Mini Kit following the manufacturer's protocol. Viral RNA was tested by the multiplex format of the RT-PCR Assay and results were confirmed by DENV E gene sequencing. No convalescent specimens were obtained from these patients to test for IgM anti-DENV seroconversion. The CDC DENV-1–4 Real-Time RT-PCR Assay was further evaluated by testing 371 archived, paired acute and convalescent serum samples obtained (2007–2011) from CDC's routine passive dengue surveillance system in Puerto Rico and 19 countries: Trinidad and Tobago, St. Martin, Venezuela, Mexico, Colombia, Brazil, Dominican Republic, India, Barbados, Samoa, Haiti, Ecuador, Costa Rica, Grenada, Honduras, Thailand, and other Netherland Antilles in accordance with a CDC IRB approved protocol. The convalescent samples, all collected at least 6 days after onset of symptoms, were tested with the IgM anti-DENV Capture Enzyme Linked Immunosorbent Assay (CDC MAC-ELISA) [48].

The following twelve organisms in the differential diagnosis of patients with a febrile illness suspected of dengue were selected for the specificity validation: West Nile Virus (WNV), yellow fever virus (YFV), St. Louis encephalitis virus (SLEV), Chikungunya virus (CHIKV), herpes simplex virus 1 and 2 (HSV-1-2), cytomegalovirus (CMV), varicella zoster virus (VZV), and bacterial organisms Leptospira, and Borrelia. These pathogens were obtained from CDC repositories.

Laboratory-grade, human endogenous interferent biomolecules were obtained and diluted in normal human serum to the following physiologically relevant pathological concentrations: bilirubin (342 µmol/L), cholesterol (13 mmol/L), hemoglobin (2 g/L), triglycerides (37 mmol/L) and genomic DNA (400 µg/100 mL). These concentrations follow guidelines published by the National Committee for Clinical Laboratory Standards (NCCLS), EP7-A2 (2005).

The CDC DENV-1–4 Real-Time RT-PCR Assay was designed to detect a wide variety of current clinically relevant strains of DENV-1, -2, -3, and -4 RNA in serum or plasma. Each probe was labeled with a distinct fluorophore-color allowing the simultaneous detection of the four DENV serotypes in a single reaction. The RT-PCR reactions can be assembled in singleplex or multiplex formats; both of these formats have been shown to provide equivalent sensitivity. The assay kit includes a positive control virus mix, which consists of heat-inactivated DENV-1 Haw, DENV-2 NGC, DENV-3 H87, and DENV-4 H241, and an assay internal control, termed Human Specimen Control (HSC). HSC consists of non-infectious cultured human cell material derived from the A549 cell line supernatant that tests positive for the presence of human RNaseP RNA (RP). RP RNA is present in most clinical samples and a positive signal confirms the integrity of the extraction reagents and the successful recovery of RNA. The sequence of primers and probes for both, HSC and RP, were obtained from the CDC Human Influenza Virus Real Time RT-PCR Diagnostic Panel (cat# KT0096).

The primers and probes used in the Assay were originally based on sequences previously published by Johnson et al., 2005 [34], however, significant modifications were introduced to increase complementarity with contemporary DENV sequences available in GenBank (Table 1). We selected 48 DENV-1, 50 DENV-2, 51 DENV-3, and 18 DENV-4 complete genome sequences as design templates from viruses isolated between 1998–2012, representing contemporary genotypes, and covering a wide geographical distribution. Figure 1 shows maximum likelihood trees representing the phylogeny of currently circulating lineages of each DENV serotype. Every primer and probe of the new CDC DENV-1–4 RT-PCR was compared to the corresponding serotype alignment and determined the frequency of every base in the oligonucleotide. Mismatches between PCR oligonucleotides and the sequence alignments were attended to as follows: degeneracy was introduced in a given nucleotide position when mismatch was 30% or greater between 2 or more strains; a base was replaced by another fixed base when mismatch was 90% or greater between 2 or more strains in a given position. International Union of Pure and Applied Chemistry (IUPAC) codes were used for all degenerate bases. The modified primer and probe sequences in the CDC DENV-1–4 Real-Time RT-PCR Assay are shown in Table 1 compared to the previously published sequences. The frequency in which each nucleotide position is matched to the representative strains of every serotype is also indicated. Additional information includes: target gene per serotype, target positions of every oligonucleotide relative to genome, and length of the amplicons. We found low probability of homodimer or heterodimer formation using OligoAnalyzer 3.1 (www.idtdna.com/analyzer) with predicted ΔG≥−9 kcal/mol between most oligonucleotide pair combination considering multiplex reaction conditions. This was confirmed experimentally running multiple oligonucleotide combinations in non-template control reactions and after the evaluation of all our clinical and analytical sample panels.

Once the Assay was designed and optimized, we compared performance and sensitivity against the real time RT-PCR assay designed by Johnson et al., 2005. In vitro transcribed RNA derived from the target regions was used as template and spiked normal human serum: 8 serial dilutions 1∶10, 5 replicates. Both assays were simultaneously run with the same samples in singleplex and multiplex formats according to each suggested protocol. The concentrations of virus RNA detected were measured in GCE/reaction. The CDC Real Time DENV-1–4 RT-PCR assay outperformed the Johnson 2005 assay detecting 10-fold less target RNA in 100% of the replicates in DENV-1 and DENV-2 singleplex format and DENV-1, -2, and -4 in multiplex format (Figure 2A and B respectively). Efficiency of amplification and overall performance was similar between both assays.

The limit of detection (LoD) was defined as the last dilution in which virus was detected in all replicates and 100% reproducibility can be guaranteed. The LoD of the assay was determined using a panel of quantified (GCE/mL) DENV stocks diluted in human serum or plasma for eight 1∶10 dilutions with 5 replicates of each dilution. Viral RNA from every replicate was extracted and tested following the package insert protocol. The RT-PCR Assay was run quantitatively to compare the LoD of different assay formats. The amount of serum sample processed for RNA extraction, the RNA elution volume, and the amount of RNA elution added per RT-PCR reaction were considered in order to calculate the viral RNA concentration per reaction and converted to units per mL of serum. Virus detection, measured in genome copy equivalents per mL of sample (GCE/mL), was compared between virus dilutions in human serum and plasma in singleplex and multiplex formats (Figures 3A and B, respectively). Viral RNA was occasionally detected in dilutions with 1×10²-1×10¹ GCE/mL; however the LoD was determined to be between 1×10⁴ and 1×10³ GCE/mL for all serotypes in both formats in serum and plasma.

Viral RNA extractions were performed using the Qiagen QIAamp DSP Viral RNA Mini Kit in the initial evaluations and compared with other manual or automated viral RNA extraction kits. The performance of each kit was measured by extracting viral RNA from a similar panel of viruses diluted in human serum. The lowest titer of RNA detection from each kit was determined as shown in Supporting Table S2. Virus concentrations at the LoD dilution (1×10³ GCE/mL) or at a 1∶10 dilution below the LoD (1×10² GCE/mL) were detected for each virus-type. These results validate comparable levels of DENV-1–4 RNA detection after viral RNA extraction with the selected methods.

The RT-PCR Assay was evaluated to determine detection of contemporary DENV strains from a broad geographical distribution at a LoD comparable to the reference panel described above. 29 DENV-1–4 strains from global locations were cultured, quantified and serial 1∶10 dilutions in human serum down to 1×10² GCE/mL were tested in triplicate. The observed LoD was similar to that obtained from the reference panel described above (Table 2).

The analytical specificity of the assay was evaluated by testing the nucleic acid extracted from the 12 organisms described in Methods, in triplicate, to determine if a positive signal was generated compared to samples spiked with DENV. No sample, except those containing DENV, produced a positive signal; therefore no cross-reaction was observed (Table 3).

In order to evaluate if DENV RNA detection is interfered by the presence human endogenous biomolecules, pathological concentrations of a selection of biomolecules were prepared in normal human serum. Each preparation was then spiked with the cultured DENV-1–4 laboratory-adapted strains mentioned above at concentrations of 1∶10 above the LoD (1×10⁴ GCE/mL), equal to the LoD (1×10³ GCE/mL), and 1∶10 below the LoD (1×10² GCE/mL). None of the tested biomolecules showed any significant interference in DENV detection (Table 4).

To evaluate possible cross-contamination of samples while performing the Assay, we used a panel of DENV-1–4 at high positive (10⁷ GCE/mL) and below LoD (5×10² GCE/mL) concentrations, respectively. Samples below LoD are composed of normal human serum spiked with DENV to a concentration just below the LoD. Eight replicates of the high positive and below LoD DENV concentrations were extracted using the Qiagen QIAamp DSP Viral RNA Mini Kit and tested in alternating series. All negative samples tested negative (32/32) and all DENV positive samples tested positive for DENV (32/32) (Table 5). Several below LoD samples produced CT values >37 that may correspond to detection of DENV RNA; however these results are not 100% reproducible.

Result reproducibility was evaluated at the CDC Dengue Branch and two external public health laboratories: the Puerto Rico Department of Health and the National Reference Laboratory in Costa Rica. Operator to operator, run to run, and site to site reproducibility was evaluated using five test panels which include negative, below LoD, low positive, and moderate positive samples. Cultured and quantified (GCE/mL) stocks of heat-inactivated laboratory strains DENV-1–4 were used in the panel. Two different operators tested each sample in the panel twice a day for at least 5 days. Each operator ran the entire panel twice. Every sample in the test panel was blinded for each operator, run, and device. Two different RNA extraction methods previously validated for the assay were used in the study. The CDC Dengue Branch Laboratory and one external site used the Qiagen QIAamp DSP Viral RNA Mini Kit and the second external site used the Roche MagNA Pure LC Total Nucleic Acid Isolation Kit following the manufacturer's protocol. Results of high reproducibility were obtained from each site and compared to determine the proportion agreement with the expected results (Table 6).

86 acute phase (0–5 days of illness: 8 DPO = 1; 12 DPO = 2; 22 DPO = 3; 29 DPO = 4; 15 DPO = 5; Avg DPO = 3.3, Std Dv = 1.2) serum samples were tested in the multiplex format of the assay and corroborated by E gene sequencing. A total of 48 samples produced positive sequence results, of which 47 were positive on the RT-PCR Assay (16 DENV-1, 11 DENV-2, 5 DENV-3 and 15 DENV-4). The single discordant sample was identified as DENV-3 by sequencing. The 38 samples with RT-PCR negative results in the Assay were also negative by sequencing. Applying sequencing as the reference method, the RT-PCR Assay had a 97.92% positive agreement and 100% negative agreement (Table 7). All samples tested positive for the RP control.

Of the 371 serum samples (0–5 days of illness: 36 DPO = 1; 56 DPO = 2; 81 DPO = 3; 115 DPO = 4; 83 DPO = 5; Avg DPO = 3.4, Std Dv = 1.2) tested in this evaluation, 102 had IgM anti-DENV seroconversion in the convalescent specimen. 100 out of these 102 samples were positive in the RT-PCR Assay using the multiplex format (98.04% positive agreement). Of the 2 discordant samples, one was positive for DENV-3 by singleplex and further confirmed by E gene sequencing. The other discordant specimen was repetitively negative by multiplex and singleplex, but DENV-3 positive by sequencing. Of the 269 samples with an IgM anti-DENV negative convalescent sample, 265 were negative by the RT-PCR Assay (98.51% negative agreement; Table 8); however, the remaining four samples were positive by sequencing (2 DENV-3 and 2 DENV-4). All serotypes were identified in this study (26 DENV-4, 29 DENV-2, 22 DENV-3 and 23 DENV-4) and all 371 samples had a RP positive control reaction.

To determine the effect of temperature variations on Assay performance, we compared detection of DENV-spiked human serum samples after multiple freeze/thaw cycles. Moderate and low positive dilutions were prepared in triplicate, frozen at −80°C for 24 hours, and subjected to five consecutive freeze/thaw cycles. Once thawed, every sample was processed according to the Assay protocol. We obtained a 100% qualitative agreement between the initial and post freeze/thaw cycle (1, 2, 3, 4, and 5) detection results (Supporting Figure S1).