Evaluation of a novel micro/nanofluidic chip platform for the detection of influenza A and B virus in patients with influenza-like illness

We introduced a novel micro/nanofluidic chip platform (MNCP), which is based on an isothermal nucleic acid amplification method. This study aimed to evaluate the MNCP method for influenza A and B viruses detecting and subtyping using throat swab samples from patients with influenza-like illness (ILI). A total of 266 throat swab samples from 266 non-repeated patients with ILI were tested for influenza A and B viruses using three methods, MNCP, a rapid influenza diagnostic test (RIDT), and real-time reverse transcription polymerase chain reaction (rRT-PCR). The results of MNCP were compared to those obtained by rRT-PCR and RIDT and the performance of MNCP was further evaluated. Compared with rRT-PCR results, the rates of sensitivity, specificity, overall concordance, and the kappa value of MNCP were 98.89%, 96.97%, 97.65%, and 0.95 for influenza A virus; 94.95%, 99.38%, 97.68%, and 0.95 for influenza B virus, respectively. Subtypes of influenza A viruses, e.g., A(H1N1)pdm09, A(H3N2), and A(not subtyped), and influenza B viruses could be distinguished in one MNCP assay within 1 h. Compared with rRT-PCR and MNCP, RIDT showed poor clinical sensitivity for influenza virus detection. This study showed MNCP is rapid, sensitive and versatile detecting system with potential for clinical application in pathogen diagnosis for patients with ILI.


Introduction
Pathogen diagnosis of influenza-like illness (ILI) has received increasing attention due to outbreaks of influenza and increased threat to human health from avian influenza (Bedford et al. 2015;Li et al. 2014;Zhu et al. 2019). Prompt and accurate detection of influenza virus is essential for patients with ILI, which helps to avoid unnecessary use of antibiotic therapy and to prevent nosocomial transmission of influenza virus (Neuzil et al. 2000;Townsend and Eiland 2006). Here, we introduced a novel micro/nanofluidic chip platform (MNCP) for influenza virus identification and its value in clinical application was evaluated by comparing MNCP with a rapid influenza diagnostic test (RIDT) and real-time reverse transcription polymerase chain reaction (rRT-PCR).

Study design and sample collection
A total of 266 throat swab samples were collected from 266 non-repeat patients with ILI according to the criteria of diagnostic and treatment protocol for influenza between Dec. 2017 andFeb. 2018 in Beijing Tsinghua Changgung Hospital (Beijing, China). The swab samples were obtained from the surface of the tonsils and the posterior pharyngeal wall by qualified personnel using nylon fiber flocked swabs. Duplicate samples were collected, one was subjected to prompt RIDT assay, the other was Open Access immediately placed in a tube containing virus transport medium (Yocon Biological Pharmacy, Beijing, China) and stored at − 80 °C for rRT-PCR and MNCP assays within 3 months (Eisfeld et al. 2014). The study was approved by the ethics committee of Beijing Tsinghua Changgung Hospital (Approval No. 17120-0-01). The flowchart of the study was shown in Fig. 1.

Design of MNCP
MNCP was developed based on the principle of nucleic acid sequence-based amplification (NASBA), which is a homogeneous and isothermal reaction (Compton 1991). In a typical NASBA process, three enzymes, avian myeloblastosis virus reverse transcriptase (AMV-RT), RNase H, and T7 RNA polymerase work together, giving the ability to exponentially amplify single-stranded RNA (Rodrıǵuez-Làzaro et al. 2004) (Fig. 2a). Molecular beacon probes specifically bind to the amplification products, resulting in fluorescence emission that is due to fluorescence resonance energy transfer and can be detected by a detector (Fig. 2b).
The general workflow of the MNCP is shown in Fig. 3. Briefly, nucleic acids were extracted from throat swab samples using TIANamp Virus DNA/RNA Kit (Tiangen Biotech, Beijing, China). The nucleic acid (8 µl) was mixed with the amplification reagents, and the mixture was loaded into the micro/nanofluidic chip, in which 24 reaction chambers were each connected to a sine-shaped infusing channel by a micro-channel. Four pairs of primers and molecular beacon probes specific for influenza A virus matrix protein (MP) gene, influenza A(H1N1) pdm09 hemagglutinin (HA) gene, influenza A(H3N2) HA gene, and influenza B virus MP gene were employed to detect subtypes of influenza viruses. The primer and beacon probe sequences were designed according to Table 1. Subtype influenza A(H1N1)pdm09 or A(H3N2) was identified if the influenza A(H1N1)pdm09 HA gene or A(H3N2) HA gene was positive. Influenza A(not subtyped) was identified as influenza A MP gene positive Fig. 1 Flowchart of the study. MNCP: micro/nanofluidic chip platform; RIDT: rapid influenza diagnostic test; rRT-PCR: real-time reverse transcription PCR plus A(H1N1)pdm09 HA gene and A(H3N2) HA gene negative. A positive control, a negative control, and an internal control were applied for each sample test as quality control. The positive control and the negative control were used to confirm that the chip was working correctly. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene was used as the internal control to monitor whether the extraction of nucleic acids from the swab samples was successful and detect whether the amplification process was valid. For each assay, the positive control, the negative control, the internal control, and the four pairs of primers and molecular beacon probes were preloaded into seven different reaction chambers, and the chip was fully enclosed in a cover to eliminate the risk of contamination (Zhang et al. 2018). Fluorescence intensity was measured twice every minute and the data were analyzed using the software that came with the RTisochip-A detector (CapitalBio, Beijing, China).

rRT-PCR
Briefly, nucleic acid extracted from the swabs was tested by rRT-PCR using a commercial kit (Bioperfectus Technologies, Jiangsu, China) accredited by China Food and Drug Administration on the Roche Cobas Z480 realtime PCR system (Roche Diagnostics, Shanghai, China) and following the manufacturer's instructions (Sun et al. 2013). The threshold cycle value of ≤ 34.7 and ≤ 34.8 was interpreted as positive for influenza A virus and influenza B virus, and the threshold cycle value > 34.7 and > 34.8 was interpreted as negative for influenza A virus and influenza B virus, respectively. A batch of rRT-PCR tests took about 2 h to complete. Viral concentration of the samples was calculated by comparing rRT-PCR results with the standard curves generated from inactivated standard influenza A and B virus following the manufacturer's instructions. Ten influenza A virus positive samples and ten influenza B virus positive samples were randomly selected for HA gene sequencing (Beijing Genomics Institute of China) to validate the accuracy of the rRT-PCR method.

Rapid influenza diagnostic test (RIDT)
Rapid influenza diagnostic test was performed using Clearview Exact Influenza A&B (ABON Biopharm Co.,  The workflow of MNCP. The steps were as follows: (1) Nucleic acid was extracted from throat swab samples. (2) A mixture including nucleic acid, basic NASBA reaction reagents, primers, and molecular beacon probes was loaded into the micro/nanofluidic chip. (3) The loaded chip was placed into the RTisochip-A detector (CapitalBio, Beijing, China). (4) RNA was amplified based on the NASBA and the fluorescence intensity data were collected and analyzed using the software. (5) Results of detecting and subtyping of influenza A and B viruses were obtained. Fluorescent curves were plotted by detection time (X-axis) against fluorescent intensity (Y-axis) for influenza A(H1N1)pdm09, A(H3N2), and influenza B virus, respectively. A batch of the MNCP tests took about 1 h. MNCP: micro/nanofluidic chip platform; NASBA: nucleic acid sequence-based amplification Table 1 The primer and beacon probe sequences Ltd, Hangzhou, China) following the manufacturer's instructions. The test strip was coated with antibodies against nucleoproteins of influenza A and B virus at different positions to distinguish antigens between influenza A and B virus (Yao et al. 2019).

Statistical analysis
Results of MNCP were compared with results of rRT-PCR and RIDT respectively, and rRT-PCR was considered the standard. Qualitative variables were summarized by their frequency distribution. The sensitivity, specificity, positive predictive values (PPVs), negative predictive values (NPVs), overall concordance rate, the kappa value, and the 95% confidence intervals of MNCP and RIDT were determined. McNemar's test and the kappa value were used to analyze the statistical difference and agreement between methods. The kappa value of 0.21-0.39, 0.60-0.79, and 0.90-0.99 indicated a minimal, moderate and almost perfect agreement between two methods, respectively (McHugh 2012). A P-value of < 0.05 was considered statistically significant. Statistical analyses were performed using SPSS 24.0 software (IBM SPSS, Chicago, IL, USA).

Baseline characteristics and the results of rRT-PCR
The baseline characteristics of the patients were shown in Table 2 (Table 3). Six of 255 samples showed opposite results between MNCP and rRT-PCR for influenza A virus detection (1, MNCP negative and rRT-PCR positive; 5, MNCP positive and rRT-PCR negative). Six of 259 samples showed opposite results for influenza B virus detection (5, MNCP negative and rRT-PCR positive; 1, MNCP positive and rRT-PCR negative) (Table 4).

Subtyping the influenza virus by MNCP
Infection of influenza virus genre and its subtyping could be achieved within 1 h with the MNCP assay using one chip for each sample. Results showed that influenza A(H1N1)pdm09, A(H3N2), A(not subtyped), and influenza B virus accounted for 27.84% (71/255),

Results of MNCP compared with RIDT
The methodological comparison results in Table 5 showed  (Table 6).

Virus concentration determined the agreement between RIDT and MNCP
As for influenza A virus detecting, 20 samples with RIDT/MNCP positive results had a median viral concentration of 2.19E+07 (10 7.34 ) copies/ml, and 69 samples with RIDT negative plus MNCP/rRT-PCR positive results had a median viral concentration of 2.69E+06 (10 6.43 ) copies/ml, indicating significantly statistical difference between the two groups (P < 0.001), as shown in Fig. 4a. As for influenza B virus detecting, 54 samples with RIDT/MNCP positive results had a median viral concentration of 1.66E+07 (10 7.22 ) copies/ml, and 40 samples with RIDT negative plus MNCP/rRT-PCR positive results had a median viral concentration of 2.75E+05 (10 5.44 ) copies/ml, with significantly statistical difference between the two groups (P < 0.001), as shown in Fig. 4b.

Discussion
The recently developed MNCP technique has been successfully used for rapid and high throughput detection of bacteria and viruses (Kaur et al. 2018), showing good sensitivity and specificity (Kim et al. 2017;Zhou QJ et al. 2014). Our study showed that compared with rRT-PCR, MNCP had good performance for influenza A and B virus detection, with the overall concordance rate, sensitivity, specificity, PPV, and NPV ≥ 95%, both the kappa values at 0.95, indicating almost perfect agreement with rRT-PCR. Subtypes of influenza A(H1N1)pdm09, A(H3N2), A(not subtyped), and influenza B virus could be distinguished simultaneously by MNCP within 1 h using one chip for each sample, which greatly reduced the workload and shortened the reporting time compared with rRT-PCR (2 h for one batch of test) (Lee et al. 2001). The amplification mixture of MNCP was loaded into different chambers of the chip, each chamber included one pair of primers and molecular beacon probes. It enabled the MNCP method to avoid the drawbacks of conventional multiplex PCR in which primer pairs significantly interfered and competed with each other in a single tube . In addition, because recombinant mutation of influenza A occurs frequently (Abed et al. 2005;, influenza virus subtyping by MNCP will be a helpful tool for clinical and viral epidemiological studies, it can even recognize new recombinant mutants at an earlier stage.
Further analysis found opposite results between MNCP and rRT-PCR assays in small numbers of samples, six samples each in influenza A and B virus detection.
According to the reagent instruction, cycle of negative threshold values for influenza A virus and influenza B virus were > 34.7/undetected, and > 34.8/undetected in rRT-PCR tests, respectively. Six samples with rRT-PCR positive/MNCP negative results for influenza virus B showed CT values were between 30.5 and 34.8. This study showed that MNCP may be more sensitive than rRT-PCR for influenza virus A detection, and less sensitive for influenza virus B detection Zhang et al. 2018).
RIDT has been used clinically for a long time; however, the poor sensitivity of RIDT causes great concerns. This study also suggested low sensitivity of RIDT, since 74 of 255 samples for influenza A virus detection and 42 of 259 samples for influenza B virus detection showed opposite results between MNCP and RIDT. Using rRT-PCR as the standard method, MNCP results were in good agreement with rRT-PCR results, with an overall concordance rate of 93.24% and 95.24% for influenza A and B virus detecting, respectively. Considering that viral concentration might affect RIDT results, we quantitatively analyzed the virus concentration of the samples and the frequency distribution of the RIDT results. Samples with a median viral concentration of 2.69E+06 copies/ml for influenza A virus and 2.75E+05 copies/ml for influenza B virus could have RIDT negative results. The study further indicated that RIDT had a poor clinical sensitivity for influenza virus detection, which was in line with a report from Germany. In the German study, only 11% of 144 PCR-positive frozen clinical specimens were positive when tested retrospectively with Binax NOW Influenza A & B test (Inverness Medical), a commercial kit similar to the one used in our study (De la Tabla et al. 2010;  Drexler et al. 2009). In consequence, a significant proportion of patients with false negative results in RIDT could lead to delayed diagnosis and treatment (Ginocchio et al. 2009). The single application of RIDT for the detection of influenza virus infection should be discouraged in favor of the clinical application of rRT-PCR and MNCP for patients with ILI (Drexler et al. 2009).
Our study also indicated that the application of RIDT to diagnose influenza A virus infection was more likely to have false negative results than to diagnose influenza B virus infection. This further suggested that the sensitivity of RIDT should be improved, especially for influenza A detection; as such testing results may mislead the clinical decision and epidemiological analysis. In recent years, the epidemic of influenza B virus has attracted much attention. Previous studies had suggested that influenza B virus could cause severe cases and fatalities as same as influenza A virus (Paul Glezen et al. 2013). It has been reported that the B/Yamagata lineage of influenza B virus was the predominant strain responsible for the For respiratory pathogen detection, poor quality specimens may induce false negative results for RT-PCR and RIDT (Drexler et al. 2009). It was worth noting that more concern should be given to the standardized sample collection. The swabs should be taken from the surface of the tonsils and the posterior pharyngeal wall, rather than from the surface of the oral cavity, to ensure that viruses can be collected (Eisfeld et al. 2014). In the MNCP detecting process, a positive control and a negative control were applied to monitor whether the chip was working correctly, while the internal control used the GAPDH gene to validate the swab sampling process. Using these controls could minimize the probability of having falsenegative results that may mislead clinical decision.
This study had some limitations.
(1) The MNCP chip was designed only for four subtypes of influenza virus. Therefore, other subtypes, such as influenza A(H1), A(H5), influenza B(Victoria) and B(Yamagata), were not included in the study. (2) The fact that MNCP reactions were in multiple chambers could increase the cost of running the assay. Ideally, both RNA extraction and amplification should be conducted using the same platform. (3) This was a single-center study and the sample population was relatively small.
In conclusion, MNCP was a homogeneous, isothermal amplification platform that can shorten the virus detection time for patients with ILI. The reaction chambers of the chip allow multiple influenza virus detection and subtyping. MNCP had the same sensitivity and specificity as rRT-PCR. MNCP was a novel, rapid, sensitive, and multipurpose detecting system that has potential clinical values in pathogen diagnosis for patients with ILI.