A six‐plex droplet digital RT‐PCR assay for seasonal influenza virus typing, subtyping, and lineage determination

Abstract Background There are two influenza A subtypes (H1 and H3) and two influenza B lineages (Victoria and Yamagata) that currently co‐circulate in humans. In this study, we report the development of a six‐plex droplet digital RT‐PCR (ddRT‐PCR) assay that can detect HA and M segments of influenza A (H1, H3, and M) and influenza B (Yamagata HA, Victoria HA, and M) viruses in a single reaction mixture. It can simultaneously detect six different nucleic acid targets in a ddRT‐PCR platform. Methods The six‐plex ddRT‐PCR used in this study is an amplitude‐based multiplex assay. The analytical performance of the assay was evaluated. Correlation with standard qRT‐PCR methodology was assessed using 55 clinical samples. Results The assay has a wide dynamic range, and it has good reproducibility within and between runs. The limit of quantification of each target in this assay ranged from 15 copies/reaction for influenza B Victoria M gene to 45 copies/reaction for influenza B Yamagata M gene. In addition, this assay can accurately quantify each of these targets in samples containing viral RNAs from two different viruses that were mixed in a highly skewed ratio. Typing, subtyping, and lineage differentiation data of 55 tested clinical respiratory specimens were found to be identical to those deduced from standard monoplex qRT‐PCR assays. Conclusions The six‐plex ddRT‐PCR test was demonstrated to be highly suitable for detecting dual influenza infection cases. This assay is expected to be a useful diagnostic tool for clinical and research use.

from pigs in the 2009 pandemic, and it replaced the classical seasonal H1N1 virus thereafter. Viruses of the four subtypes/lineages circulate widely in humans and 0.5%-3.0% of influenza patients are dually infected. 2,3 Such dual infection cases are reported, but not necessarily confined to, children, young adults, pregnant women, and immunocompromised patients. 2,[4][5][6] Quantitative RT-PCR (qRT-PCR) is a sensitive and specific nucleic acid test for detecting influenza virus. 7 Viral loads determined by qRT-PCR assays are suggested to be useful markers for assessing disease severity and for predicting clinical outcomes. [8][9][10] However, quantification of nucleic acid targets in standard qRT-PCR assays heavily relies on the quality of external standards and the relative signal-to-noise ratio. 11,12 Thus, the performance, reproducibility, and amplification efficiency of a standard qRT-PCR-based assay can vary greatly between different laboratories. The recent development of droplet digital PCR (ddPCR) provides an alternative solution to overcome this potential hurdle. The ddRT-PCR assay uses microfluidics and emulsion chemistries to generate about 20 000 partitions or droplets per reaction. 13 With each of these emulsified droplets containing approximately ≤ 1 copy of studied targets, quantification of copy numbers can be reliably calculated using Poisson statistics. 14 Thus, the ddPCR approach does not require the use of a standard curve for quantification.
Multiplex assays have several advantages over standard monoplex assays, in terms of reducing reagent cost, sample consumption, hands-on processing time, accumulated pipetting inaccuracy etc. 15,16 There are real-time multiplex qRT-PCR assays for typing and subtyping seasonal influenza viruses. Some can even detect influenza B virus and determine influenza B viral lineage, 17,18 but none of these assays allows simultaneous typing and subtyping/lineage differentiation of all 4 seasonal influenza viruses in a single qRT-PCR reaction. Further, ddRT-PCR does not require a standard curve for quantification and it is less susceptible to PCR inhibitors. 19 Here, we report an efficient six-plex ddRT-PCR that can enhance influenza surveillance and epidemiologic studies thereby informing immunization policies, control strategies, and outbreak responses. The assay allows influenza virus typing and subtype/lineage determination in a single reaction.

| Primers and probes
Primers and probes targeting human influenza viruses (A: H1pdm09 or H3; B: Vic or Yam) were designed based on sequences available in ing criteria: (a) the last five bases at the 3′-end of a primer should be a perfect match to its target; (b) the last ten bases at the 3′-end of a primer should not have more than one mismatch to its target; and (c) the maximum number of mismatches between a primer and its target should not be more than two. A sequence from the database was considered as closely or perfectly matched to the probe sequence if not more than one mismatch was found. All primers and probes were synthesized commercially (Integrated DNA Technologies). All the probes were labeled with a 5′-fluorophore (FAM or HEX), a 3′-Iowa Black FQ quencher and an internal ZEN quencher.

| Six-plex ddRT-PCR
For developing a six-plex ddRT-PCR, reactions were prepared using a one-step ddRT-PCR Advanced Kit for Probes (Bio-Rad) as in-  ddRT-PCR reactions were kept temporarily at 4°C, and reaction signals were captured by a droplet reader (QX200™ system, Bio-Rad).
All reactions were required to yield a minimum of 10 000 droplets per reaction before downstream analyses. Data generated from the droplet reader were analyzed by a program designed for this platform (QuantaSoftTM Pro, version 1.0 BioRad). The gating strategy was set and optimized manually at the beginning of this study. The optimized gating strategy was used in all subsequent experiments.
No-template control reaction was included in each run.

| Evaluation of six-plex ddRT-PCR
To test cross-reactivity, six-plex ddRT-PCR were set-up under different conditions: single and double positive using viral RNAs from virus stocks. The dynamic range of the assay was established using ten-fold dilutions of viral RNAs from virus stocks. Each dilution was run in replicates on three different days, in which triplicates were performed on day 1 to show intra-assay variability with a total of five replicates to show the inter-assay variability.
Based on the dynamic range results, limit of quantification (LoQ) was set to be the lowest concentration of each targets that could be quantified with CV ≤ 25%. Sixteen replicates were run over three different days for the determination of LoQ of each target (triplicates on day 1, five replicates on day 2, and eight replicates on day 3). The limit of blank (LoB) was set to be the 95th percentile of positive droplets in reactions using the ten negative control samples, the samples were analyzed in duplicate to get twenty sets of data in total.
To compare the performance of this assay between single and double positive reactions, mixed samples were prepared using viral RNAs from two different virus stocks in various ratios. For one target, copies per reaction were very low (<99) or low (99-250), whereas for the other target, copies per reaction were high (3000-13 000) or very high (13 000-70 000). The error was calculated by subtracting the quantitative data for a double positive reaction from that for a single positive reaction. The mean percentage error was calculated using the data in three independent runs.  tested. These samples were previously typed and classified by qRT-PCR assays using protocols as previously described. [21][22][23][24] Briefly, viral RNA was extracted using QIAamp viral RNA mini kit (Qiagen) according to the manufacturer's instruction. qRT-PCR was performed using Qiagen One-Step RT-PCR kit (Qiagen) and was conducted by thermal cycler (ViiA7 Real-time PCR system, Thermo Fisher). RNA samples positive for H1pdm09 (N = 11), H3 (N = 12), Vic (N = 12), or Yam (N = 10), together with negative control samples (N = 10), were tested by the six-plex ddRT-PCR assay in a blinded format. Ct values for these samples in the above qRT-PCR assays and the data generated from the ddRT-PCR were analyzed in Prism8.

| In silico analyses of primer and probe sequences
In this study, we developed a multiplex assay for detecting seasonal  (Table 1). These primer and probe sequences were highly specific to contemporary influenza virus sequences (2009-2019) and >97% of the studied influenza sequences should react with our sequence designs based on the criteria used in our study (see Materials and Methods).
As the commercial ddRT-PCR reader has only two channels for signal detection, we adopted the amplitude multiplexing technique to detect our targets. 15 The primer-probe sets were pre-optimized to different concentrations so that the detected signals would form distinct clusters in a 2-dimensional amplitude multiplexing plot (Figure 1).

| Performance of the six-plex assay for influenza virus detection
We first conducted the assay using viral RNA extracted from virus cultures. The extracted RNA from a particular strain was tested either alone or in mixture with another viral RNA in different combinations in order to make the studied reactions become single or double positive. As shown in Figure 1A, reactions containing viral RNA from different viruses yielded distinct signal cluster patterns. In double positive reactions, no cross-reactivity was observed in all kinds of combinations ( Figure 1B). It should be noted that we intended to use highly diluted RNA samples to achieve ≤ 1 copy per droplet for ddRT-PCR. The numbers of double positive droplets were therefore expected to be low. If more concentrated RNA mixtures were used in this assay, the number of clusters would have increased ( Figure S1). The sample containing the H1pdm09 and Yam had no positive cluster higher than 7000 arbitrary units in the HEX channel, but there was a positive cluster centered at 5000 arbitrary units in the HEX channel ( Figure 1B). The sample with H3 and Yam RNAs had positive clusters from 5500 to 7000 arbitrary units in the HEX channel. The same interpretation rules indicated above were applied to reactions having higher concentrations of target samples ( Figure S1B).
We determined the dynamic range of the assay by using 10-fold serially diluted RNA samples. The assay had a dynamic range of at least four orders of magnitude (Figure 2), which is typical for digital PCR assays with ~10 000 droplet/reaction. 25 Results from this range of dilutions showed a linear relationship (R 2 ≥ 0.981). We also conducted multiple replicates to determine the intra-assay variability (N = 3) and inter-assay variability (N = 5). All tests had a CV value of less than 25%, with more dilute RNA samples tending to have a higher CV value as expected ( Table 2). We further conducted tests on sixteen replicates to determine the limit of quantification (LoQ) of the assay ( Table 3). The CV values of LoQ of this assay were all less than 25%, meeting the recommended standard for microbial detection. 26 The LoQ values of this assay (copies/reaction) for our targets In order to confirm that signals from one virus do not significantly interfere with those from another virus, we mixed two different viral RNA samples in various ratios and tested these mixtures using the ddRT-PCR assay. The quantitative results were compared with those deduced from control reactions with RNA from a single virus. As shown in Table 4, the quantitative data of one target generated from mixed RNA samples were in concordance with the expected values (mean absolute percentage error ranged from 2.3% to 20.9%). These results indicated that this assay is suitable for identifying samples with dual influenza virus infections.
Using RNA samples extracted from ten irrelevant respiratory clinical specimens as negative controls, we determined the number of false positive droplets in negative reactions (ie, limit of blank, LoB). The 95th percentile of positive droplets for FAM and HEX signals in a negative reaction (N = 20) were found to be 8 and 6, respectively. Thus, experimental reactions with numbers of positive droplets of less than 8 in FAM and 6 in HEX were considered as negative. Reactions with number of positive droplets above LoB, but below LoQ, are classified as "positive but not quantifiable."

| Evaluation of ddRT-PCR assay using clinical specimens
To evaluate the performance of this assay for clinical diagnosis, we tested 45 retrospective influenza-positive and 10 control RNA samples extracted from nasopharyngeal swabs. These RNA samples were previously tested by influenza typing and subtyping using qRT-PCR (See Methods and Materials), and double influenza infection was not detected in these samples. All typing and subtyping/lineage differentiation results generated from the ddRT-PCR assay agreed with those deduced from the qRT-PCR assays. We further compared results generated by the qRT-PCR assay (Ct values) with those generated by ddRT-PCR assays (copies per reaction). These two sets of data were highly correlated (Figure 3, R 2 > 0.938 for all targets). Overall, our results show that the ddRT-PCR assay is a robust test for simultaneous typing, subtyping, and lineage determination of human influenza types A and B viruses.

Results from recent clinical studies indicate that influenza virus load
can be a marker for disease severity. 9 Influenza A and B viruses are often detected and quantified by qRT-PCR in modern clinical settings. 7 However, while this diagnostic approach is highly robust, qRT-PCR is sensitive to inhibitors in clinical samples and absolute quantification of its target(s) is highly reliant on accuracy of the standard curve(s), resulting in significant inter-laboratory variations. 27 Hence, there is need to develop an accurate molecular test that does not require a standard curve for quantification.
In this study, we have developed a novel multiplex ddRT-PCR for typing, subtyping, and lineage differentiation of human influenza viruses in a single reaction. As the QX200™ reader used in this study can recognize only two fluorophores simultaneously, 15 we adopted an amplitude multiplexing approach to allow detection of six different targets in one reaction. The double-quenched probe can provide flexibility on the probe length and reduce the background of detection. 28 The six-plex ddRT-PCR, compared with the monoplex qRT-PCR, can reduce reagent cost, handling time, and manpower and allow sample conservation. 15,16 Our results show that our targets positive signals display distinct patterns in a 2D analysis system (Figure 1). One should note that the complexity of clusters mainly depends on the concentration of targets in double positive reactions ( Figure 1B and Figure S1B).

Previous works from others showed that ddRT-PCR technology
can detect up to sixteen clusters with a maximum of four different targets in a single reaction. 15 Thus, to the best of our knowledge, our ddRT-PCR study is the first that shows formation of distinct clusters with the capacity to detect up to six targets in a typical positive reaction ( Figure 1B), with a broad range of linearity ( Figure 2) and good reproducibility within and between runs ( The error was calculated by subtracting the copy number of a target in a double positive reaction from the copy number of the same target in a single positive reaction. Data are presented as mean absolute percentages of three runs.