3.1. A genome-wide SNP map of WNV evidences substantial variation
To test the hypothesis that single nucleotide polymorphisms (SNPs) are present in recent WNV isolates and that SNPs may be present in any single one PCR primer binding site that is used for diagnosis, genbank was queried. 1,628 SNPs out of 11,029 nucleotide positions were found by BLAST search of the prototypical lineage I strain AF196835 (). Of those 82 sequences the medium number of SNPs per genome was 34 with a range of 0-499 SNPs. The SNPs were distributed evenly across the genome, except for the 3′ and 5′ UTR regions (), which exhibited a relative paucity of SNPs. The E orf, too, contained numerous SNPs, which were distributed across the length of the orf (). This is a conservative estimate, since only genbank entries classified as containing complete WNV genome sequences were included in our alignment and of those only n=82 sequences with less than 500 mismatches (). For instance, strain AF196835 shows more than 2000 vis-à-vis strains belonging to other lineages such as lineage II, AY532665; lineage III, AY765264; lineage IV, AY277251; and lineage V, DQ256376). Moreover, strains classified as lineage I-B (D00246), show more than 1000 mismatches when aligned against the AF196835 sequence.
Figure 1 SNP distribution across the WNV genome. Panel A-D SNP positions are indicated by a black bar. A, distribution across the entire genome; B, distribution across the most 5′ 1,529 nucleotides; C, distribution across the E orf; D, distribution across (more ...)
To gain a more detailed view at the SNP distribution, the nucleotide distances between any two adjacent SNPs were calculated (). The mean distance between any two SNPs in the WNV genome was 6.75 nucleotide with a 95CI of 6.46 - 7.05 and a range of 0 - 54 nucleotides. Since a QPCR amplicon averages 90-200 nucleotides, it can be expected that diagnostic tests based upon only a single PCR assay will, sooner or later, fail because of SNPs in either one of the primer or the probe binding sites.
The 5′UTR up to nucleotide 420 and the 3′ UTR up to nucleotide −513 contained large inter-SNP regions, indicative of sequence conservation. This was expected since these regions are conserved among WNV and other flaviviruses in general. Both St. Louis encephalitis virus (SLE) and Japanese encephalitis (JEV) share extensive sequence identity in the UTR regions of up to 81% for SLE isolate BeAn 247377 (accession # EF158067) and up to 78% for JEV isolate K01-JN (accession# AY965851) for the 5′ UTR, respectively. Hence, for the purpose of differential WNV detection, primers that target the UTR regions are expected to have low specificity.
In the U.S. the possibility of misclassifying an infection as SLE rather than WNV raises concerns. Additionally, viral UTRs should not be selected as targets for diagnostic PCR probes due to the fact that UTR regions exhibit extensive and stable secondary structures. For the purpose of WNV detection, these may not be efficiently transcribed by reverse transcriptase or not unfold completely during the short denaturation step of recent “fast” real-time QPCR protocols. Hence, primers that target the UTR regions are expected to have low sensitivity. In sum, there are few absolutely conserved regions in the WNV genome and targeting them may not yield the best assay. It can be expected that there exist WNV sequence variants that escape detection with current single primer-pair-based assays.
3.2. A novel set of WNV primers validated by annealing temperature profiling
One approach to overcome potential false negative results caused by single mutations within any viral genome is to use multiple primers against the same virus. To test this hypothesis a panel of novel WNV-specific QPCR primers were developed with equal sensitivity as well as an integrated, high throughput assay format that reduce greatly false negative rates.
The chief concern when using multiple primers against the same target is that primers may exhibit varying annealing temperatures and thus cannot be combined on a single plate in a 96 or 384 well format. “PrimeTime” is based on Primer3 (Rozen, Skaletsky, 2000
) to design multiple primers against WNV that share the same melting temperature (Tm
) and are compatible with each other. The initial primer design strategy which was used imposed no limitations as to the location of a given primer, it only required strict adherence to Tm
and universal primer design parameters (Papin et al., 2004a
). This approach identified multiple primers across the WNV genome ().
Primer position, name, orientation and sequence based upon NY99-flamingo382-99, complete genome (AF196835). SNP positions are underlined.
QPCR optimization typically evaluates primer concentration, Mg2+
concentration and annealing temperature. First, the influence of the primer concentration was evaluated. All primers yielded single QPCR products with comparable efficiency (Keff.
) at 150 nM final concentration, consistent with our prior experience using real-time QPCR arrays of >80 primers for human herpesviruses (Dittmer, 2003
; Dittmer et al., 2005
; Fakhari, Dittmer, 2002
; Hilscher et al., 2005
). Adjusting reagent conditions, such as Mg2+
concentration individually for each primer pair is impractical as it limits high throughput, automated analyses with multiple primers per target. It was not necessary here, since all primer pairs designed according to our parameters performed equally at a given Mg2+
concentration (data not shown). Therefore, the reagents provided in the commercial 2x SYBR mix were used. It canhypothesized, however, that one reason for decreased primer pair efficiency was a suboptimal annealing temperature under our QPCR conditions.
predictions (Cantor, Schimmel, 1980
) as provided by most primer design programs are not accurate, since the exact salt concentrations and buffer composition of most commercial QPCR reagents are not available. To test this hypothesis the gradient feature of the Opticon2 (MJR research/Biorad Inc.) real-time QPCR unit was used, which combines gradient PCR with the capability to record product accumulation quantitatively in real-time in the standard 96 well format. Hence, an annealing temperature dependent change in performance can be quantified on the same plate without post-PCR processing. Next a gradient annealing step of 55°C to 65°C for 15 sec. followed by a 68°C 15 sec. extension step, followed by fluorescence measurements, followed by a 94°C 10 sec denaturing phase was employed. Fluorescence in each well was always measured at the same temperature. Hence, the resulting CT values reflect primarily differences in annealing efficiency. Other factors include primer extension efficiencies, the formation of primer dimmers, the formation of primer hairpins, template stability (including GC content), and template secondary structure (such as hairpin formation), among others, but not temperature dependencies during the extension phase or dye intercalation. Using automation, the optimal annealing temperature for sixteen primers pairs across six temperature points were determined at the same time. Most combinations of primers with a Primer3-predicted Tm
of 60°C showed amplification at 60°C, but that temperature profiles varied considerably among individual primer pairs. , panel A exemplifies a very efficient primer pair, as indicated by a median CTmedian
= 10 with little variation across the annealing temperature range. , panel B shows a less efficient primer pair, as indicated by CTmedian
= 13.5, but with likewise little variation across the annealing temperature range. , panel C shows a still less efficient primer pair, as indicated by CTmedian
= 23.5. It showed little variation across a wide annealing temperature range. By contrast, , panel D shows a primer pair with a strong annealing temperature dependence as evidenced by a continuously changing CT between the lowest and highest annealing temperature. , panel E shows a primer pair with a higher annealing temperature dependence. It failed to function above 64°C. In sum, many primer combinations functioned equally efficiently within ±2°C of their predicted Tm
. Thus assays using multiple primer pairs against the same target, WNV, can be combined on the same instrument and 96 or 384 well plate.
Figure 2 Annealing temperature dependence of qPCR for individual primers pair combinations. The vertical axis is reversed as lower CTs indicate a higher target abundance or, as all reactions received the same amount of input target a higher qPCR efficiency. The (more ...)
After 40 cycles, all products for a given primer pair yielded the same size single band on an agarose gel and the same single peak melting temperatures (data not shown). This was expected since the primer design program incorporated a BLAST (Altschul et al., 1997
) step, which excluded primers that bound anywhere other than the intended targets (Dittmer et al., 2005
). This evidence confirms that a change in annealing temperature did not yield to a change in product, but reflected purely the efficiency of primer annealing.
Absolute quantitation using an external standard is required for viral copy number determination, as amplification efficiencies between primer pairs can vary considerably. Of note, the two worst performing primer pairs () also yielded considerably longer qPCR products of 2651 bp and 2282 bp, respectively, which under fast qPCR cycling protocols as are used here (extension ≤30sec) are associated with lower qPCR efficiency. A linear relationship (p≤0.005, by regression analysis) between amplicon length and median CT () was evident for amplicon lengths above 500 bp. For amplicon length below 500 bp, performance in qPCR was not dependent on amplicon length. The degree of annealing temperature variation was not dependent on amplicon length, since large interquartile differences were observed for the smallest (42 bp) as well as for larger (486 bp, 1042 bp, 2281 bp) amplicons. This shows that amplicon length can limit assay sensitivity for the detection of WNV and that an amplicon size of 500 bp or less is associated with higher sensitivity.
Figure 3 Amplicon length dependence of qPCR. Boxplot of CT values at different annealing temperature (vertical axis) for primer pairs of different length (shown on the horizontal axis). Boxframe indicates the 1 and 3 quartiles, the solid bar the median and whiskers (more ...)
3.3. An optimized, equal amplicon size, equal annealing temperature, and equal efficiency set of WNV primers for viral load determination and phylogenetic analysis
To address the issue of amplicon length dependent changes in primer efficiency, i.e. assay sensitivity, a subset of real-time qPCR primer pairs were chosen that all yielded a 400bp product (), and that exhibit a temperature invariant annealing temperature profile around 60±5°C (data not shown) under rapid qPCR cycling conditions. An amplicon length of 400 bp was chosen because (a) these can be easily purified by commercial, high-throughput resin-based methods, while smaller amplicons such as the typical ≤ 90bp TaqMan™ amplicons do not bind efficiently to DNA affinity resins (Qiagen, and Promega product information and unpublished observation); and (b) direct sequencing of a 400 bp PCR product yields enough information for strain typing and phylogenic tree construction with reasonable boostrap values, while the sequence information contained within <100 bp amplicons does not (Whitby et al., 2003
Figure 4 Genome-wide qPCR. Picture of an ethidium bromide stained agarose gel yielding same size amplicons (A) or combinations to yield largers amplicons (B). The primer names are indicated above the gel. Note that the primer names do not indicate the map position, (more ...)
None of the primers showed substantial cross reactivity to human cDNA (data not shown). More importantly, even though one individual primer pair may anneal theoretically somewhere on the human genome, the pair combinations shown in do not yield any product when used on human cDNA. The reason is that rapid real-time qPCR conditions as in do not produce efficiently amplification products larger than 1000 bp and, based on BLAST search, no two primers used in annealed on the human genome within even 10,000 bp of each other (data not shown).
By using any three primer pairs from this collection instead of a single primer triplicate repetition assay format, the false negative rate due to SNPs in one of the primer binding sites can be reduced 6 fold, while still employing multiple independent QPCR reactions per sample. In a decision model based on maximal sensitivity, such as to exclude any possibly WNV positive blood samples, a single positive amplification would cause a sample to be considered positive. In a decision model weighted towards specificity, such as to exclude samples that may contain a related flavivirus, three positive amplifications would cause a sample to be coded positive.
Should the phylogenic information gained from 400 nt not suffice for accurate typing, this set of WNV QPCR primers can be combined to yield larger PCR products that cover 80% of the WNV genome () without change in annealing temperature. This approach was used to determine the complete sequence of WNV OK03, which has been deposited in genbank (accession number EU155484). This allows for the sequence determination of virtually the entire viral genome using the same reagents and annealing temperature as used for WNV viral load determination. The only change that is necessary is an increase in the extension time from 30 sec to 120 sec. For completeness, the primer combinations indicated in can be used to amplify the entire viral genome using a common annealing temperature of 60 °C and an extension time of 60 seconds. PCRs using first primer (position 1) and the last primer (position 10,876) require an annealing temperature of 55°C.
Typically viral load assays are conducted as replicate measurements of the same assay/ primer pair. Here, the same number of reactions are run but instead of one primer pair multiple different, but annealing temperatures and Keff.
-compatible primer pairs, such as introduced in are used. This will decreases the false negative rate due to nucleotide variation in the target sequences of any primer pair. This is particularly applicable to viral pathogens with seasonal patterns of incidence, such as WNV, since from one season to the next multiple adoptive mutations can be expected (Davis et al., 2003
; Davis et al., 2005
3.4. PCR volume optimization contributes significantly to assay sensitivity
As the next step in optimizing high sensitivity, high specificity real-time qPCR assays for WNV, qPCR volume and input to reaction ratio were evaluated. As expected more sample per qPCR correlated with higher assay sensitivity, which here is indicated by lower cycle threshold (CT) numbers (). However, the relationship was not linear, but rather followed a power decay function. In conclusion, larger reaction volumes ultimately lead to less efficient heat transfer and thus less efficient PCR. The ratio of input sample to total reaction volume was more important than overall reaction volume, such that PCRs comprising only 20% of input sample were much more sensitive (as indicated by lower CT volumes) than PCRs comprising 42% input sample, even though the total amount of sample was higher in the latter. At 42%, the cDNA reaction buffer components interfere with the QPCR buffer components leading to decreased PCR efficiency.
Figure 5 Volume dependence of qPCR. (A) Shown on the horizontal axis is the amount of input cDNA sample and on the vertical axis the mean CT of four replicate measurements after real-time qPCR. Lower CT corresponds to better detection on a 2log scale. (B) Shown (more ...)
Quadruplicate qPCR yielded standard deviations (SD) of ≤ 1 CT unit, or ≤ 2-fold, regardless of sample volume. This establishes that quadruplicate measurements suffice to distinguish two-fold differences in WNV viral load. In this case, too, a power decay function best fit the data, demonstrating that lower sample volumes were associated with reduced assay accuracy. The loss in assay accuracy was likely due to increased pipetting error at low volumes. In sum, an input sample volume of 20 μl in a total qPCR volume of 100 μl yielded the most sensitive and most accurate real-time qPCR assay for WNV.
3.5. Maximum assay sensitivity by using high volume reverse transcription
Three commercial kits used to isolate WNV RNA were compared: Viral RNA (Qiagen), HighSensitivity (Qiagen), and viral RNA (Zymogen). All three kits were column-based. It can be presumed that the column binding capacity to be in excess of any amount of viral RNA that is found in bodily fluid samples such as plasma, whole blood or CSF. For tissue samples (bird, biopsy), previously a modified Triazol™ (Sigma) procedure (Papin et al., 2004b
) was used, which has a higher capacity than these column-based kits and eliminates the risk of overloading as associated with resin/column-based RNA isolation. Furthermore, a prior reverse transcription assay (Papin et al., 2004b
) was compared to the new high-capacity cDNA kit (Applied Biosystems). The goal was to arrive at a method that maximizes the number of viral genome cDNA equivalents per well of the QPCR reaction.
The main differences between the different isolation methods were the maximal input volume and optimal output volumes, which result in the dilution of RNA. traces the effects of volume dilution starting from a theoretical concentration of 1000 copies/ml. Combining high sensitivity/high input RNA isolation with a high capacity cDNA kit yielded the most copies (50) as input for qPCR and the most number of possible replicates (nine+one). If less than 1000 μl input material was available, a low elution volume combined with customary low volume RT reaction yielded the most copies (12) as input for qPCR while still allowing triplicate measurements (three+one). Of note the high-sensitivity/high yield RNA-isolation/reverse transcription combination resulted in a relative amplification of target copy number during the reverse transcription step. Assuming 1 copy/μl input (or 103 copies/ml titer), after RNA isolation/reverse transcription high yield/high capacity RT concentrated the RNA to 2.5 copies/μl, while all other combinations diluted the input down to as much as 0.35 copies/μl in the worst case.