An integrated microfluidic system for rapid detection and multiple subtyping of influenza A viruses by using glycan-coated magnetic beads and RT-PCR

Kao-Mai Shen a, Narayana Murthy Sabbavarapu b, Chien-Yu Fu a, Jia-Tsrong Jan b, Jen-Ren Wang c, Shang-Cheng Hung *bd and Gwo-Bin Lee *aef
aDepartment of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, 30013 Taiwan. E-mail: gwobin@pme.nthu.edu.tw; Fax: +886 3 5742495; Tel: +886 3 5715131 Ext. 33765
bGenomics Research Center, Academia Sinica, Taipei 11529, Taiwan. E-mail: schung@gate.sinica.edu.tw; Tel: +886 2 27871279
cDepartment of Medical Laboratory Science and Biotechnology, College of Medicine, National Cheng Kung University, Tainan 701, Taiwan
dDepartment of Applied Science, National Taitung University, 369, Section 2, University Road, Taitung 95092, Taiwan
eInstitute of Biomedical Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan
fInstitute of NanoEngineering and Microsystems, National Tsing Hua University, Hsinchu 30013, Taiwan

Received 15th December 2018 , Accepted 22nd February 2019

First published on 22nd February 2019


The influenza A (InfA) virus, which poses a significant global public health threat, is routinely classified into “subtypes” based on viral hemagglutinin (HA) and neuraminidase (NA) antigens. Because there are nearly 200 viral subtypes, current diagnostic approaches require multiplexing or array systems to cover various subtypes of HA and NA. A microfluidic chip featuring a HA × NA array was consequently developed herein for diagnosis and subtyping of InfA viruses via the use of glycan-coated magnetic beads followed by reverse transcription (RT) polymerase chain reaction (PCR). Up to 12 InfA subtypes were simultaneously detected in an automated fashion in less than 100 minutes on this microfluidic platform, representing a significant improvement in analysis speed compared to benchtop RT-PCR and chip-based microarray systems. The limits of detection of the RT-PCR assays ranged from 40 to 3000 copy numbers for the different subtypes of InfA viruses, around two orders of magnitude higher than in previous studies using microfluidic technologies. In summary, the array-type microfluidic chip system provides a rapid, sensitive, and fully automated approach for detection and multiple subtyping of InfA.


Introduction

Influenza, more commonly known as the “flu”, is a global public health threat caused by ribonucleic acid (RNA) viruses. According to the World Health Organization, influenza epidemics affect 3 to 5 million individuals annually and lead to 250[thin space (1/6-em)]000 to 500[thin space (1/6-em)]000 deaths worldwide each year.1 Influenza viruses are classified into four types: A, B, C, and D.2 The influenza A (InfA) virus can be divided into different subtypes based on combinations of hemagglutinin (HA) and neuraminidase (NA) antigens on the viral surface. At present, there are 18 and 11 known types of HA and NA antigens, respectively; this means that nearly 200 InfA viral subtypes could exist,3 presenting a diagnostic challenge.

Currently, the subtypes of InfA that most commonly infect humans are H1N1, H2N2, and H3N2. However, increasingly more and more evidence has shown that various subtypes of avian influenza viruses (e.g. H5N1, H5N6, H6N1, H7N2, H7N3, H7N7, H7N9, H9N2, H10N7 and H10N8) can cross the species barrier to infect humans,4 and high mortality rates have been reported for humans infected with avian influenza H5N1 (449/850 cases, 53%), H5N6 (9/15 cases, 60%), and H7N9 (314/791 cases).4 Therefore, it is important to rapidly carry out accurate diagnostic procedures to determine the infecting subtypes such that the best treatment, prevention and control strategies can be applied.

For influenza diagnostic and subtyping methods, viral culture has been the gold standard for decades; however, it is poorly suited for timely clinical management because of its lengthy protocols and labor-intensive nature. Furthermore, well-trained technicians and specialized equipment are both required.5 Serological testing is only useful for surveillance purposes and is usually not recommended for routine patient diagnosis and clinical management because it is characterized by limited sensitivity and specificity.6 Alternatively, rapid influenza diagnostic tests (RIDTs) are commonly used for influenza diagnosis because they exhibit the benefits of high specificity (>90%) and rapid analysis time (<30 min) while less technical skills and specialized equipment are required. However, their poor sensitivity (10–70%) compared with viral culture and reverse transcription polymerase chain reaction (RT-PCR) methods and inability to differentiate subtypes of InfA may limit their widespread applicability.6

Recently, molecular diagnostic techniques, such as RT-PCR, have gained popularity in many influenza diagnostic labs because they are highly sensitive, specific, versatile, and relatively rapid (3–4 h); however, these approaches are nevertheless labor-intensive and require relatively expensive reagents, well-trained technicians, and specialized equipment.3,6 Another molecular diagnostic approach, the chip-based DNA microarray, is based on DNA amplicon hybridization and has been used to quickly diagnosis and subtype multiple InfA viruses simultaneously.7–10 In addition, several microarrays used to detect respiratory viruses have been produced commercially, including the INFINITI® FLU A-sH1N1 assay (AutoGenomics, USA) and the United States Food and Drug Administration-cleared VRNAT assay (Verigene®, USA).11 However, these assays cannot be fully automated since the viral RNA extraction and RT-PCR steps still must be undertaken manually. Hence, there is a critical need for InfA detection and subtyping devices that are rapid, simple, sensitive, cheap, and precise.3

Recently, glycans terminating in sialic acid (N-acetylneuraminic acid) have been reported to serve as useful probes for capturing influenza virus particles due to the glycan-binding specificity of the HA antigen on the viral surface.12 The binding affinity between glycans and viral HA proteins is commonly assessed with the equilibrium dissociation constant (KD), which is in the nano- to micromolar range.13–16 Glycan arrays,13,17,18 glycan-bead arrays,12 surface plasmon resonance sensors,14,15,19–21 field effect transistors,22 colorimetric assays,23 impedimetric glycan-based biosensors,24 and waveguide-mode sensors25 have all exploited glycan-virus binding properties. While the reliability and applicability of these assays must still be verified and compared to the traditional diagnostic methods, glycans nevertheless show great potential to function as probes for isolating influenza viruses.

Recently, lab-on-a-chip (LOC) technologies (miniature systems that integrate multiple laboratory functions into a single chip via micro-electro-mechanical systems [MEMS] and microfluidic technology) have been customized to feature interconnected fluidic microchannel networks, microvalves, micromixers, micropumps, reaction chambers, and detectors that can undertake many laborious benchtop protocols without human intervention or external equipment.26,27 Because of the benefits of integrated microfluidic systems (e.g., capacity for automation, high accuracy, shorter reaction times, decreased reagent consumption, and high throughput), they have been widely used in a variety of biomedical fields over the past two decades and have replaced conventional benchtop equipment in many laboratories.

Influenza viruses can be rapidly detected by 1) specific antibodies,28,29 2) nucleic acid hybridization,30–32 3) aptamers,33,34 and 4) amplification assays (i.e., RT-PCR)35 in microfluidic systems. However, some of these approaches can only identify InfA and influenza B viruses (InfB) at present, and not their individual subtypes. Even the other approaches could detect 2–3 subtypes,32 but they still require complicated processes for the sample pretreatment (e.g., virus isolation and lysis, RNA extraction and reverse transcription steps); this deficiency has substantially limited the applicability of microfluidic devices for influenza diagnostics. One commercially available, microfluidic-based microarray system known as VereFlu™ (Veredus Laboratories company, Singapore) does allow for automated, specific, rapid (<3 h), and simultaneous identification of different influenza virus strains in a single platform;11 unfortunately, viral RNAs must still be extracted manually. Hence, there is a critical demand for a fully-automated microfluidic system for 1) influenza diagnosis and 2) subtyping of influenza viruses.

In this work, we have developed an integrated microfluidic system that combined the advantages of microfluidics and RT-PCR in order to automatically perform the entire influenza virus detection and subtyping processes (including sample treatment, RT-PCR, and fluorescence detection). Based on the high specificity between influenza viruses and glycans, we used glycan-coated magnetic beads as probes to first capture all influenza viruses in a sample. The viral typing (InfA and InfB) and arrayed HA × NA subtyping (up to 4 HA subtypes and 3 NA subtypes) with one-step RT-PCR utilizing different combinations of both universal virus primers and HA- and NA-specific primers in the arrayed reaction chambers of the microfluidic chip. With this approach, up to 12 InfA influenza subtypes could be simultaneously detected in an automated fashion in less than 100 minutes on this microfluidic platform.

Materials and methods

2.1 Glycans and magnetic bead-coating

In order to initially capture all influenza viruses in samples in the developed microfluidic system, glycans were used as a probe by surface-coating them onto magnetic beads. Three kinds of sialylated N-glycans (SCH-42, SNM-01-139, and DJR-03-99) with N-acetylneuraminic acid (Neu5Ac) α(2 → 6)-linked to galactose (Gal) were prepared which made use of N-phthaloyl-protected sialyl imidate as a donor and a suitably protected lactosamine building block as an acceptor,36 and tested, and their structures are shown in Table S3. In the synthesis of the sialylated glycans, the lactosamine module was used as the repeating unit for chain elongation after the assembly of sialylated trisaccharide cores prepared by exclusive α-stereoselective sialylation. Upon successful formation of trisaccharide, it was further transformed into the reactive donor and glycosyalated with the linker attached lactosamine acceptor stereoselectively to yield fully protected sialylated N-glycans, which were then subjected to the functional group interconversions to secure the final compounds. For coating glycans onto magnetic beads, glycans were biotin-modified and conjugated to streptavidin-modified magnetic beads by the streptavidin–biotin interaction (KD = 10−15 M). Two kinds of magnetic beads (stock concentrations = 7–9 × 109 beads per mL, 1 μm in diameter, Dynabeads MyOne™ streptavidin C1 [Cat. #: 65001] and T1 [Cat. #: 65601], Thermo Fisher Scientific, USA) were tested to investigate their virus capture efficiencies. To prepare the glycan-coated magnetic beads, each glycan (20 μL of 100 μM) was conjugated with each of the C1 and T1 magnetic beads in a total volume of 100 μL (7–9 × 108 beads) on a wheeling rotator (Intelli-Mixer RM-2 L, ELMI Ltd, Latvia; 25 rpm) at room temperature for 1 hour. Subsequently, deionized distilled water (ddH2O) was used to wash the bead complex to remove unbound constituents and resuspend the glycan-coated magnetic beads in a final volume of 100 μL. Ultimately, the glycan SNM-01-139 and the C1 Dynabeads MyOne™ were found to have the highest capture efficiency for the influenza virus (Fig. S2 and S3), and SNM-01-139-coated C1 beads were used for all experiments described below.

2.2 Chip design and fabrication

The 84 mm (length) × 59 mm (width) × 10 mm (depth) microfluidic chip (Fig. 1a and b) consisted of one thick-film polydimethylsiloxane (PDMS, Sylgad 184A/B, Dow Corning, USA) air control layer, one thin-film PDMS liquid channel layer, and a glass substrate (G-Tech Optoelectronics, Taiwan) for sealing the microfluidic chip. Briefly, these PDMS layers were fabricated by a standard soft lithography process, and the PDMS layers and glass substrate were bonded together by oxygen plasma treatment.37 Several microfluidic devices were integrated into the microfluidic chip, including consecutive, pneumatically-driven micropumps for liquid transport (a modification of our previous work),37 a pneumatically-driven, open-type micromixer for incubation of samples and beads,38 pneumatically-driven and normally-closed microvalves for liquid control,39 air inlets, a waste outlet, air microchannels, liquid microchannels, loading chambers for fluid storage, overflow chambers for excess liquid storage, and RT-PCR chambers.
image file: c8lc01369a-f1.tif
Fig. 1 (a) A photograph and (b) an exploded view of the integrated microfluidic chip. Gray, blue, red, and green colors indicate the first air control layer, the second air control layer, the liquid channel layer, and the glass substrate, respectively. Arrayed reaction chambers contained primer sets for amplifying specific regions of the HA and NA genes such that the RT-PCR-derived signal output could be used for viral subtyping.

The loading chambers included one wash buffer loading chamber and 12 RT-PCR reagent loading chambers. The latter 12 included chambers for positive controls (1 ng purified viral RNA), negative controls (ddH2O), the temperature measurement, InfA typing, InfB typing, HA subtyping (H1, H3, H5, and H7), and NA subtyping (N1, N2, and N9). These HA×NA arrayed chambers permitted the identification of up to 12 subtypes of InfA viruses on-chip, and the array-type microfluidic chip would be incorporated with a custom-made control and an optical detection system (described in detail below and in our previous works)30,40 which could automatically execute the entire process of virus detection and multiple subtyping and detect the resultant signals, respectively. Methods for characterization of chip performance are described in ESI 4.1.

2.3 Working principle of the diagnostic assay

The whole process for influenza virus detection and subtyping using glycan-coated beads and on-chip, one-step RT-PCR is schematically shown in Fig. 2. First, glycan-coated beads, the virus sample, RT-PCR reagents, and wash buffer were pre-loaded into the corresponding chambers (Fig. 2a). Next, viruses were isolated using glycan-coated beads while gently mixing in the micromixer (Fig. 2b). Then, the wash buffer was pumped in to remove the unbound wastes, and magnetic bead–virus complexes were collected by applying an external magnet (Fig. 2c). Afterwards, bead-captured virus complexes were resuspended in wash buffer (Fig. 2d), and viral RNA was released by a 5 min thermolysis achieved by placing a thermoelectric (TE) cooler underneath the chip (Fig. 2e). Afterwards, the supernatant containing the viral RNA was transported to the RT-PCR reaction chamber (Fig. 2f). RT-PCR reagents were then pumped into the RT-PCR reaction chambers, and a 40-cycle RT-PCR was carried out with the TE cooler (Fig. 2g). Finally, RT-PCR signals were detected using an optical detection module (Fig. 2h). The total analysis time of this integrated microfluidic system was <100 min, and detailed experimental protocols of the steps listed above is described in Table S1.
image file: c8lc01369a-f2.tif
Fig. 2 A schematic diagram of the influenza virus detection process. (a) Loading of glycan-coated magnetic beads, virus samples, wash buffer, and RT-PCR reagents. (b) Incubation of beads and viruses by activating the open-type micromixer. (c) Washing away of unbound particles. (d) Resuspension of bead-captured virus complexes. (e) Thermal lysis of viruses with a thermoelectric (TE) cooler. (f) Transport of supernatant and PCR reagents into the reaction chambers. (g) Amplification of specific genes for HA and NA by RT-PCR. (h) Detection of fluorescence signals.

2.4 Custom-made control system

In order to automatically execute the entire process of influenza virus detection and multiple subtyping, including virus purification, virus thermolysis, liquid transport, one-step RT-PCR and optical detection, we integrated our microfluidic chip with a custom-made control system, which was reported in our previous works.30,40 The custom-made system consisted of an air compressor (UN-90 V, Uni-Crown, Taiwan), several electro-magnetic valves (EMVs; S070M-5BG-32, SMC, Japan), a temperature control module (USB4718, Advantech, Taiwan) equipped with a TE cooler (TEC1-127.10, Tande, Taiwan), and a thermocouple (SD-947, Reed Instruments, USA) for viral thermolysis and RT-PCR. An optical detection module equipped with a photomultiplier tube (PMT; R928, Hamamatsu, Japan) and a laser source (MBL5, CNI, China) was used for signal detection of the final RT-PCR products.

2.5 Influenza strains

Five different subtypes of InfA – one strain of InfA/H1N1 (105 M049), one strain of InfA/H3N2 (16028588), one candidate vaccine strain of InfA/H5N1 (Vietnam/1194/2004 NIBRG-14), one strain of InfA/H5N2 (Taiwan/duck/30-2/2005), and one candidate vaccine strain of InfA/H7N9 (Shanghai/2/2013 IDCDC-RG32A) and two strains of InfB (105M044 [Victoria strain] and 17M50121 [Yamagata strain]) were provided by the Clinical Virology Laboratory, National Cheng Kung University Hospital (NCKUH, Taiwan) and the Genomics Research Center of Academia Sinica (Taiwan). The initial titers of influenza virus stocks were quantified by a plaque assay (yielding the number of plaque forming units [PFU]),41 an HA assay (yielding the number of hemagglutinin units [HAU]),42 or by directly quantifying viral RNA copy numbers.

For the latter analysis, we constructed plasmid replicas of the InfA and InfB genomes following the user guides for the TOPO® TA cloning® kit (Cat. #: K450002, Invitrogen, USA) and generated a threshold cycle (Ct)-based real-time PCR standard curve using the primers and assays from our previous work (Wang et al. 2012).31 Then, RNA samples were extracted from the virus stocks using the PureLink™ viral RNA/DNA mini kit (Cat. #: 12280050, Invitrogen, USA), and viral copy numbers were calculated with real-time PCR in comparison with the standard curves (Fig. S1). All operations were carried out in accordance with the safety guidelines set forth by the NCKUH and the Genomics Research Center (Academia Sinica), and the initial viral concentrations determined by the aforementioned approaches can be found in Table S2.

2.6 One-step RT-PCR

One-step RT-PCR was carried out with a commercial kit (KK4660, KAPA SYBR® FAST one-step qRT-PCR kit, MERCK, Germany) in the arrayed RT-PCR reaction chambers. The RT-PCR reagents comprised 10 μL of KAPA SYBR FAST qPCR master mix (2×), 0.4 μL of deoxyuridine triphosphate (dUTP, 10 μM), 0.4 μL of KAPA RT mix (50×), 0.8 μL of each forward/reverse primer pairs for InfA/H1 (15 μM), InfA/H3 (10 μM), InfA/H5 (10 μM), InfA/H7 (10 μM), InfA/N1 (5 μM), InfA/N2 (20 μM), InfA/N9 (10 μM),43 universal InfA (10 μM)44 and universal InfB (10 μM)31 in a final volume of 12 μL. The sequences of the primers (synthesized by Mission Biotech, Taiwan) can be found in Table S4, and their specificities were verified by gel electrophoresis banding patterns (Fig. S6).

For the RT-PCR step, 12 μL of the aforementioned one-step RT-PCR reaction reagents were mixed with 8 μL of sample containing viral RNA extracted from bead-captured virus complexes, and thermocycling was performed as follows: 42 °C for 5 min to synthesize complementary DNA (cDNA), 95 °C for 5 min to denature the double-stranded DNA (dsDNA), and 40 cycles of denaturation at 95 °C for 5 s followed by annealing and extension at 60 °C for 30 s. The integrated microfluidic system and a real-time PCR system (StepOnePlus™ Real-Time PCR System, Thermo Fisher Scientific, USA) were utilized for RT-PCR, and a PCR system (ProFlex™ PCR System, Thermo Fisher Scientific, USA) was utilized for testing the specificity of InfA primers. Before the on-chip RT-PCR process, reagents were covered with 50 μL of mineral oil (Cat. #: 0008042475, Sigma-Aldrich, USA) to hinder liquid evaporation.

The primers used for both negative and positive controls were universal InfA (10 μM) or universal InfB (10 μM) forward/reverse primer pairs (the sequences of the primers can be found in Table S4). The RT-PCR products were detected optically (via their SYBR green fluorescence) on the aforementioned optical detection module, and 2% agarose (CA97062-250, VWR Life Science, USA) gels with 100 bp DNA ladders (Cat. #: DM001-R500, GeneDireX, USA) were electrophoresed with the same samples and stained with ethidium bromide (Cat. #: 15585011, Invitrogen, USA) to serve as a comparison.

2.7 Optimization of mixing conditions

For increasing the efficiency of virus capture, we optimized the mixing conditions of the micromixer by modifying the following parameters: applied pressure, operating frequency, and mixing time. First, 10 μL of SNM-01-139-coated beads (7–9 × 109 beads per mL) were mixed with 10 μL of the influenza virus stock (100-fold dilution of 4.25 × 105 PFU μL−1 of InfA/H1N1) in a final volume of 100 μL in the open-type micromixer. Different combinations of applied negative gauge pressures (0, −6.7, −13.3, −40.0, −66.7, or −93.3 kPa), operating frequencies (0.0, 0.5, 1.0, 2.0, 3.0, or 4.0 Hz), and incubation times (0, 0.5, 1.0, 2.5, 5.0, 10.0, or 30.0 min) were all tested at a positive gauge pressure (13.3 kPa), and analyzed by RT-PCR-derived Ct value.

2.8 Calculation of capture rates

In order to evaluate the capture rates of SNM-01-139-coated beads for influenza viruses, 10 μL of SNM-01-139-coated beads (7–9 × 109 beads per mL) were mixed with 10 μL aliquots of 100-fold dilutions of influenza virus stocks in a final volume of 100 μL in the open-type micromixer. The supernatant removed from the virus capture and washing steps, as well as the bead-captured virus complexes, were collected, and their viral RNA was extracted by using the PureLink™ viral RNA/DNA mini Kit as described above. Next, viral copy numbers were estimated by real-time RT-PCR as described above. Finally, the capture rates were calculated as the number of viral copy numbers in the bead complexes over the total number of virus copies measured in the bead complex and supernatant fractions (×100), which was shown as follows (eqn (1)).
 
image file: c8lc01369a-t1.tif(1)

2.9 Sensitivity tests

In order to evaluate the limits of detection (LODs) of this developed microfluidic system, 10 μL of SNM-01-139-coated beads (7–9 × 109 beads per mL) were mixed with each 10 μL of 10-fold serial dilutions of influenza virus stocks in a final volume of 100 μL in the open-type micromixer. The logarithmic dilutions from 0 to 6 (for InfA/H1N1, InfA/H3N2, InfB/Victoria, and InfB/Yamagata) or 0 to 5 (for InfA/H5N1, InfA/H5N2, and InfA/H7N9) resulted in virus concentrations ranging from 4.3 × 10−2 to 4.3 × 105 PFU mL−1 (6.4 × 10−5 to 64 HAU mL−1 and 8 to 8.1 × 106 copy numbers per μL) of InfA/H1N1, 8.5 × 10−3 to 8.5 × 104 PFU mL−1 (3.2 × 10−5 to 32 HAU mL−1 and 4 to 4.1 × 106 copy numbers per μL) of InfA/H3N2, 5 to 5 × 106 PFU mL−1 (6 to 6 × 105 copy numbers per μL) of InfA/H5N1, 10 to 107 PFU mL−1 (4 to 4 × 105 copy numbers per μL) of InfA/H5N2, 20 to 2 × 107 PFU mL−1 (30 to 3 × 106 copy numbers per μL) of InfA/H7N9, 3.5 × 10−1 to 3.5 × 106 PFU mL−1 (6.4 × 10−5 to 64 HAU mL−1 and 2 to 2 × 106 copy numbers per μL) of InfB/Victoria, and 7.8 × 10−1 to 7.8 × 106 PFU mL−1 (1.3 × 10−4 to 1.3 × 102 HAU mL−1 and 5 to 5 × 106 copy numbers per μL) of InfB/Yamagata, respectively. Note that the virus stocks with a volume of 10 μL per reaction were used in this experiment. After completing the on-chip processes, viral RNAs in the supernatants were amplified by on-chip RT-PCR with the corresponding typing and subtyping primers. Then, the amplified signals were quantified using the custom-made optical detection module equipped with the PMT and confirmed by gel electrophoresis.

2.10 Viral subtyping

In order to test whether different subtypes of influenza viruses could be identified using this developed microfluidic system featuring HA- and NA-specific subtyping primers, 10 μL of SNM-01-139-coated beads (7–9 × 109 beads per mL) were mixed with 10 μL of each of the influenza virus stocks in a final volume of 100 μL by activating the open-type micromixer. The subsequent analysis was identical to that shown in section 2.9.

Results and discussion

3.1 Characterization of the micropump and micromixer

The performance of the microfluidic device, including its pumping rate, pumping precision across consecutive micropumps, and mixing index of the micromixer, was characterized (Fig. 3 and 4), and the developed micropump and micromixer could be used for sample/reagent transport and incubation/mixing, respectively. The pumping rate of pump 1 (“1” in Fig. 3a) was maximized at −66.7 kPa and 2 Hz (Fig. 3b) while the pumping rate of pump 2 (“2” in Fig. S3a) was the highest at −53.3 kPa and 2 Hz. However, both micropumps were relatively unstable at −66.7 kPa and 2 Hz. Therefore, we ultimately chose −53.3 kPa and 2 Hz as the operating conditions of the micropumps for transporting the RT-PCR reagents from the RT-PCR reagent loading chambers into the RT-PCR reaction chambers in the subsequent experiments; the pumping rates of pumps 1 and 2 were 143.8 ± 1.7 and 142.1 ± 1.7 μL min−1 (std. dev; n = 3). Fig. 3c shows that the micropumps (numbered 1 to 10 in Fig. 3a) could transport liquid uniformly from the right micromixer to the middle reaction chambers at applied negative gauge pressures below −40 kPa; the associated fluid pumping volume was around 8.0–8.5 μL per round of transport. Because of the instability of the left-most micropumps (unnumbered), the liquid transported from the left-most micropumps was typically transported into the overflow chambers. Finally, we chose −53.3 kPa as the operating condition of the consecutive micropumps for transporting the supernatant containing the viral RNA from the micromixer into the RT-PCR reaction chambers in subsequent experiments. The results of mixing index shows that applying operating frequencies of 0.5, 1.0, and 2.0, Hz could complete the mixing within 4, 2, and 1 s, respectively (Fig. 4a–c). The pneumatic, open-type micromixer used herein was characterized by a higher mixing efficiency than the pneumatic close-type micromixer,37 which required 10 s for complete mixing at 1 Hz. A possible reason for this phenomenon might result from the different mixing efficiency between vortex flow (open-type) and chaotic flow (closed-type). These results also indicate that the applied negative gauge pressure does not appreciably affect the mixing efficiency of the open-type micromixer.
image file: c8lc01369a-f3.tif
Fig. 3 Characterization of the micropumps. (a) A schematic of the chip featuring the consecutive micropumps. (b) The pumping rates of the micropumps (labeled 1 and 2 in panel a) at different applied negative gauge pressures (−6.7, −13.3, −26.7, −40.0, −53.3, or −66.7 kPa) and different operating frequencies (0.5, 1.0, and 2.0 Hz). (c) The pumping precision of each micropump (see panel a.) of the consecutive micropumps at different applied negative gauge pressures (−40.0, −53.3, −66.7, or −80.0 kPa). In panels b and c, error bars represent standard deviation (n = 3).

image file: c8lc01369a-f4.tif
Fig. 4 Mixing index measurements at different operating frequencies (0.5, 1.0 Hz, or 2.0 Hz) at different applied negative gauge pressures (−13.3, −26.7, −40.0, −53.3, or −66.7 kPa) at a constant applied positive gauge pressure (13.3 kPa). The mixing index at 0.5 (a; 4 s for complete mixing), 1 (b; 2 s for complete mixing), and 2 (c; 1 s for complete mixing) Hz. Error bars represent standard deviation (n = 3).

3.2 Optimization of mixing conditions for capturing influenza viruses

Virus capture efficiencies did not differ across applied negative gauge pressures (Fig. 5a) for InfA/H1N1, except for when the samples were not mixed (0 kPa). These results were similar to those of the mixing index of the open-type micromixer (Fig. 4), indicating that the applied negative gauge pressures were not a major driver for mixing efficiency. According to this result, we chose a relatively low negative gauge pressure, −13.3 kPa, for all subsequent experiments since it may not pose a strong shear force. Virus capture efficiencies achieved a maximum value at operating frequencies of 2 and 3 Hz for InfA/H1N1 (Fig. 5b), dropping at an operating frequency of 4 Hz. All subsequent experiments were consequently carried out using an open-type micromixer with an operating frequency of 2 Hz. Virus capture efficiencies were experimentally found to increase with time and saturated after 10 min (i.e., complete occurred within 10 min); this time was therefore used in all subsequent experiments and is similar to that of our prior works.28,30,31,33,34 This suggests that the glycan-coated beads are likely to have a similar affinity for influenza viruses to the probes used in these prior studies.
image file: c8lc01369a-f5.tif
Fig. 5 Optimization of mixing conditions for capturing influenza viruses (n = 3; error bars represent standard deviation). (a) The capture efficiency of the micromixer at different applied negative gauge pressures (0, −6.7, −13.3, −40.0, −66.7, or −93.3 kPa) for InfA/H1N1. (b) The capture efficiency of the micromixer for InfA/H1N1at different operating frequencies (0, 0.5, 1.0, 2.0, 3.0, or 4.0 Hz). (c) The capture efficiency of the micromixer for InfA/H1N1at different capturing times (0, 0.5, 1.0, 2.5, 5.0, 10.0, or 30.0 min). (d) The capture rates of SNM-01-139-coated beads for different influenza viruses on the microfluidic system.

3.3 Capture rates of the microfluidic system for influenza viruses

Under the optimal mixing conditions (−13.3 kPa, 2 Hz, and 10 min), the capture rates of SNM-01-139-coated beads for influenza viruses were explored (Fig. 5d), and rates for InfA/H1N1, InfA/H3N2, InfA/H5N1, InfA/H5N2, InfA/H7N9, InfB/Victoria, and InfB/Yamagata, were found to be 59 ± 9% (standard deviation; n = 3 for all experiments), 72 ± 4%, 18 ± 2%, 63 ± 4%, 26 ± 4%, 89 ± 2%, and 76 ± 7%, respectively. The notably lower capture rates for InfA/H5N1 and InfA/H7N9 might be due to the strategy to produce reassortant viruses of the A/Puerto Rico/8/1934 (PR8) strain, which contributed to the low HA production of vaccine viruses e.g., NIBRG-14 and IDCDC-RG32A.45,46 According to these results, we deem these capture rates to be suitable for diagnostics.

3.4 Limits of detection (LODs)

The LODs of universal InfA, InfA/H1, and InfA/N1 primers were experimentally found to be 4.25 × 10−2, 4.25 × 10−1, and 4.25 × 10−1 PFU mL−1 (10 μL/reaction) for InfA/H1N1, respectively, with optical signals of 2.5±0.6, 2.5±0.1, and 3.6±1.8 V, respectively (Fig. 6a and Fig. S5). Fig. 6b shows the electropherograms of the amplified RT-PCR products as a comparison. The LODs of universal InfA, InfA/H3, and InfA/N2 primers were experimentally found to be 8.50 × 10−2, 8.50 × 10−3, and 8.50 × 10−2 PFU mL−1 (10 μL per reaction) for InfA/H3N2 (Fig. S5a), respectively, with corresponding optical signals of 4.6 ± 1.1, 2.6 ± 0.3, and 2.3 ± 0.4, respectively, using the optical detection module. The LODs of universal InfA, InfA/H5 and InfA/N1 primers were all 50 PFU mL−1 (10 μL per reaction; Fig. S5c) for InfA/H5N1, with optical signals of 3.8 ± 0.3, 3.6 ± 0.3, and 3.3 ± 0.4 V, respectively. The LODs of universal InfA, InfA/H5 and InfA/N2 primers were measured to be 1.00 × 101, 1.00 × 102 and 1.00 × 101 PFU mL−1 (10 μL per reaction; Fig. S5e) for InfA/H5N2, and optical signals of 2.9 ± 0.4, 3.8 ± 0.8, and 2.9 ± 0.2 V, respectively, were measured. Those of InfA/H7N9 (Fig. S5g) were 20, 20, and 200 PFU mL−1 (10 μL per reaction), respectively, with optical signals of 3.1 ± 0.3, 2.1 ± 0.1, and 3.5 ± 0.6 V, respectively. The LODs of the universal InfB primer were 3.5 and 7.8 × 10−1 PFU mL−1 (10 μL per reaction; Fig. S5i) for InfB/Victoria and InfB/Yamagata, respectively, with optical signals of 4.5 and 2.5 V, respectively. Gel images confirm these results (see Fig. S5b, d, f, h, and j, for InfA/H3N2, InfA/H5N1, InfA/H5N2, InfA/H7N9, and InfB/Victoria+InfB/Yamagata respectively.)
image file: c8lc01369a-f6.tif
Fig. 6 The limits of detection (LODs) of the microfluidic system with SNM-01-139-coated beads and subtyping primers for InfA/H1N1. (a) The LODs detected using the optical detection module equipped with a photomultiplier tube (PMT). Error bars represent standard deviation (n = 3). (b) The LODs confirmed by electrophoresis. Lane M = 100 bp DNA ladder, lane P = positive control (1 ng viral RNA), lane N = negative control (ddH2O), and Lane “No virus” = PBS + SNM-01-139-coated beads only.

According to these results, the LODs of this microfluidic system for InfA/H1N1, InfA/H3N2, InfA/H5N1, InfA/H5N2, InfA/H7N9, InfB/Victoria, and InfB/Yamagata were determined to be 4.3 × 10−1, 8.5 × 10−2 PFU mL−1, 50, 100, 200, 3.5, and 7.8 × 10−1 PFU mL−1, respectively. Furthermore, the LODs were also quantified by viral copy numbers, and LODs of 810, 408, 622, 382, 2973, 176, and 50 copies per reaction were calculated for InfA/H1N1, InfA/H3N2, InfA/H5N1, InfA/H5N2, InfA/H7N9, InfB/Victoria, and InfB/Yamagata, respectively (Table 1). For LODs expressed in HAUs, as well as a comparison against LODs of prior works, please see Table 2. Briefly, our system was characterized by a lower or comparable LOD for each of the target influenza viruses when compared to pre-existing technologies; it could be related to higher probe affinity, better mixing performance, and/or improved sample quality. Moreover, the LODs of the developed system were also comparable to the commercial LOC device VereFlu™ (200 copies per reaction for pandemic InfA/H1N1, 800 copies per reaction for seasonal InfA/H1N1, and 100 copies per reaction for InfA/H3N2 and InfB viruses). Therefore, the sensitivity of this developed microfluidic system should be sufficient for influenza diagnosis.

Table 1 The limits of detection (LODs) of the specific primers for different influenza viruses, as represented by viral copy numbers (n = 3 except for InfB [denoted by #]). “—” = not detected
Virus strains Primers
InfA InfB H1 H3 H5 H7 N1 N2 N9
H1N1 81 810 810
H3N2 408 41 408
H5N1 622 622 622
H5N2 38 382 38
H7N9 297 297 2973
InfB/Vic 176#
InfB/Yam 50#


Table 2 Comparison of the limits of detection (LODs) of the microfluidic influenza diagnostic chip developed herein versus LODs of other devices
Virus type Technology LOD (HAU) References
H1N1 Glycan bead+RT-PCR chip 6.4 × 10−4 Herein
H1N1 Aptamer-based sandwich assay 3.2 34
H1N1 Antibody-based sandwich assay 3.1 × 10−2 28
H1N1 Antibody-based sandwich assay 7 × 10−3 47
H1N1 Aptamer-based sandwich assay 3.2 × 10−2 33
H1N1 Aptamer-based RT-PCR assay 6.4 × 10−3 48
H1N1 Antibody-based RT-PCR assay 6.4 × 10−1 48
H1N1 DNA-probe RT-PCR 1.3 × 10−2 30
H3N2 Glycan bead + RT-PCR chip 3.2 × 10−4 Herein
H3N2 Aptamer-based sandwich assay 3.2 34
H3N2 Antibody-based sandwich assay 5 × 10−4 29
H3N2 DNA-probe RT-PCR 1.3 × 10−2 30
infB-Yamagata Glycan bead + RT-PCR chip 1.2 × 10−4 Herein
infB-Victoria Glycan bead + RT-PCR chip 6.4 × 10−4 Herein
infB Aptamer-based sandwich assay 3.2 × 10−2 34
infB Antibody-based sandwich assay 1.3 × 10−1 28
infB DNA-probe RT-PCR 1.3 × 10−3 30


3.5 Subtyping of influenza viruses

The developed system successfully identified the InfA/H1N1 subtype with extremely high optical signal output (Fig. 7a): 7.4 ± 1.1, 7.0 ± 1.4, 5.7 ± 1.8, and 5.4 ± 1.3 V for the positive control (330 bp), InfA (330 bp), InfA/H1 (85 bp), and InfA/N1 (126 bp) RT-PCR assays, respectively (Fig. 7b). Similarly, optical signals for the InfA/H3N2 subtype (Fig. 7c) were 6.8 ± 0.1, 6.6 ± 0.7, 5.8 ± 0.4, and 5.3 ± 1.2 V in the positive control (330 bp), InfA (330 bp), InfA/H3 (118 bp), and InfA/N2 (122 bp) RT-PCR assays, respectively (Fig. 7d).
image file: c8lc01369a-f7.tif
Fig. 7 The subtyping results of various influenza viruses using either the developed microfluidic system with the optical detection module equipped with photomultiplier tubes (PMT; left-most panels; error bars represent standard deviation [n = 3]) or electropherograms (right-most panels). (a) and (b) InfA/H1N1; (c) and (d) InfA/H3N2; (e) and (f) InfA/H5N1; (g) and (h) InfA/H5N2; (i) and (j) InfA/H7N9; (k) and (l) InfB/Victoria; (m) and (n) the subtyping result of InfB/Yamagata. Lane M = 100 bp DNA ladder, lane P = positive control (1 ng viral RNA), lane N = negative control (ddH2O), lane A = InfA, and lane B = InfB.

For InfA/H5N1 (Fig. 7e), high optical signals of 5.9 ± 0.3, 5.7 ± 0.2, 5.2±0.4, and 5.6 ± 0.5 V were measured in the positive control (330 bp), InfA (330 bp), InfA/H5 (90 bp), and InfA/N1 (126 bp) RT-PCR assays, respectively (Fig. 7f). For InfA/H5N2 (Fig. 7g), optical signals of 6.7 ± 0.7, 6.6 ± 1.0, 5.7 ± 1.4, and 6.4 ± 1.1 V were measured in the positive control (330 bp), InfA (330 bp), InfA/H5 (90 bp), and InfA/N2 (122 bp) RT-PCR assays, respectively (Fig. 7h). For InfA/H7N9 (Fig. 7i), optical signals of 6.6 ± 0.7, 6.4 ± 0.6, 4.6 ± 0.5, and 4.9 ± 0.3 V were measured in the positive control (330 bp), InfA (330 bp), InfA/H7 (91 bp), and InfA/N9 (77 bp) RT-PCR assays, respectively (Fig. 7j). For InfB/Victoria (Fig. 7k), optical signals of 5.7 ± 0.4 and 5.8 ± 0.9 V were measured in the positive control (170 bp) and InfB (170 bp) RT-PCR assays, respectively (Fig. 7l). For InfB/Yamagata (Fig. 7m), optical signals of 6.0 ± 0.5 and 5.8 ± 0.4 V were measured in the positive control (170 bp) and InfB (170 bp) RT-PCR assays, respectively (Fig. 7n). In all cases, experiments were repeated thrice, and each run from sample loading to data report required about 100 min (sample-to-answer assay), versus around 3 hours for VereFlu™ and the other previous works, which require completed processes on sample pretreatment and well-trained personnel to handle it.11,32

Conclusions

In this work, we developed a novel, integrated microfluidic system capable of automating the process of virus purification, virus thermolysis, liquid transport, one-step RT-PCR, and fluorescence signal detection for diagnosis of multiple subtypes of influenza viruses within 100 min. Under optimized operating conditions, the capture rates of the microfluidic system for different influenza viruses were generally over 50%. Additionally, the LODs of the specific RT-PCR assays ranged from ∼40 to 3000 copies, comparable to, or better than, those reported previously. InfA/H1N1, InfA/H3N2, InfA/H5N1, InfA/H5N2, InfA/H7N9, InfB/Victoria, and InfB/Yamagata could all be detected by this arrayed microfluidic system, which could be readily modified to feature probes capable of binding a wide array of pathogens with HA×NA arrayed combinations. In summary, the influenza microfluidic diagnostic chip provides a rapid, sensitive, simple, sample-to-answer, and fully-automatic approach for detection of multiple influenza virus subtypes and is consequently suitable for point-of-care diagnostics and surveillance in the near future.

Conflicts of interest

There are no conflicts of interest to declare

Acknowledgements

The authors would like to acknowledge financial support from the Ministry of Science and Technology (MOST) of Taiwan (MOST 106-2221-E-007-001 and MOST 105-2119-M-007-009 to GBL). Partial financial support from the “Higher Education Sprout Project” of Taiwan's Ministry of Education (Grant No. 107Q2713E1) is also greatly appreciated.

References

  1. L. A. Reperant, F. M. Moesker and A. D. Osterhaus, ERJ Open Res., 2016, 2, 00013–2016 CrossRef PubMed.
  2. L. Ferguson, A. K. Olivier, S. Genova, W. B. Epperson, D. R. Smith, L. Schneider, K. Barton, K. McCuan, R. J. Webby and X. F. Wan, J. Virol., 2016, 90, 5636–5642 CrossRef CAS PubMed.
  3. S. V. Vemula, J. Zhao, J. Liu, X. Wang, S. Biswas and I. Hewlett, Viruses, 2016, 8, 96 CrossRef PubMed.
  4. C. M. Bui, A. A. Chughtai, D. C. Adam and C. R. MacIntyre, Arch. Public Health, 2017, 75, 15 CrossRef PubMed.
  5. D. E. Dwyer, D. W. Smith, M. G. Catton and I. G. Barr, Med. J. Aust., 2006, 185(10 Suppl), S48–S53 Search PubMed.
  6. F. Lopez-Medrano, E. Cordero, J. Gavalda, J. M. Cruzado, M. A. Marcos, P. Perez-Romero, N. Sabe, M. A. Gomez-Bravo, J. F. Delgado, E. Cabral and J. Carratala, Enferm. Infecc. Microbiol. Clin., 2013, 31, e521–e526 Search PubMed.
  7. X. Han, X. Lin, B. Liu, Y. Hou, J. Huang, S. Wu, J. Liu, L. Mei, G. Jia and Q. Zhu, J. Virol. Methods, 2008, 152, 117–121 CrossRef CAS PubMed.
  8. Y. Huang, H. Tang, S. Duffy, Y. Hong, S. Norman, M. Ghosh, J. He, M. Bose, K. J. Henrickson, J. Fan, A. J. Kraft, W. G. Weisburg and E. L. Mather, J. Clin. Microbiol., 2009, 47, 390–396 CrossRef CAS PubMed.
  9. X. Kang, Y. Li, H. Sun, W. Wu, H. Liu, F. Lin, C. Qing, G. Chang, Q. Zhu, W. Chen and Y. Yang, Arch. Virol., 2010, 155, 55–61 CrossRef CAS PubMed.
  10. X. Li, X. Qi, L. Miao, Y. Wang, F. Liu, H. Gu, S. Lu, Y. Yang and F. Liu, Diagn. Microbiol. Infect. Dis., 2009, 65, 261–270 CrossRef CAS PubMed.
  11. J. Teo, P. Di Pietro, F. San Biagio, M. Capozzoli, Y. M. Deng, I. Barr, N. Caldwell, K. L. Ong, M. Sato, R. Tan and R. Lin, Arch. Virol., 2011, 156, 1371–1378 CrossRef CAS PubMed.
  12. M. Cohen, C. J. Fisher, M. L. Huang, L. L. Lindsay, M. Plancarte, W. M. Boyce, K. Godula and P. Gagneux, Virology, 2016, 493, 128–135 CrossRef CAS PubMed.
  13. H. Y. Liao, C. H. Hsu, S. C. Wang, C. H. Liang, H. Y. Yen, C. Y. Su, C. H. Chen, J. T. Jan, C. T. Ren, C. H. Chen, T. J. Cheng, C. Y. Wu and C. H. Wong, J. Am. Chem. Soc., 2010, 132, 14849–14856 CrossRef CAS PubMed.
  14. C. F. Mandenius, R. Wang, A. Alden, G. Bergstrom, S. Thebault, C. Lutsch and S. Ohlson, Anal. Chim. Acta, 2008, 623, 66–75 CrossRef CAS PubMed.
  15. E. Suenaga, H. Mizuno and K. K. Penmetcha, Biosens. Bioelectron., 2012, 32, 195–201 CrossRef CAS PubMed.
  16. T. Takahashi, S. Kawagishi, M. Masuda and T. Suzuki, Glycoconjugate J., 2013, 30, 709–716 CrossRef CAS PubMed.
  17. S. Gulati, Y. Lasanajak, D. F. Smith, R. D. Cummings and G. M. Air, Cancer Biomarkers, 2014, 14, 43–53 Search PubMed.
  18. J. Stevens, O. Blixt, L. Glaser, J. K. Taubenberger, P. Palese, J. C. Paulson and I. A. Wilson, J. Mol. Biol., 2006, 355, 1143–1155 CrossRef CAS PubMed.
  19. P. Critchley and N. J. Dimmock, Bioorg. Med. Chem., 2004, 12, 2773–2780 CrossRef CAS PubMed.
  20. K. I. Hidari, S. Shimada, Y. Suzuki and T. Suzuki, Glycoconjugate J., 2007, 24, 583–590 CrossRef CAS PubMed.
  21. E. Suenaga, H. Mizuno and P. K. R. Kumar, Virulence, 2012, 3, 464–470 CrossRef PubMed.
  22. S. Hideshima, H. Hinou, D. Ebihara, R. Sato, S. Kuroiwa, T. Nakanishi, S. I. Nishimura and T. Osaka, Anal. Chem., 2013, 85, 5641–5644 CrossRef CAS PubMed.
  23. C. Lee, M. A. Gaston, A. A. Weiss and P. Zhang, Biosens. Bioelectron., 2013, 42, 236–241 CrossRef CAS PubMed.
  24. A. Hushegyi, D. Pihikova, T. Bertok, V. Adam, R. Kizek and J. Tkac, Biosens. Bioelectron., 2016, 79, 644–649 CrossRef CAS PubMed.
  25. S. C. B. Gopinath, K. Awazu, M. Fujimaki and K. Shimizu, Acta Biomater., 2013, 9, 5080–5087 CrossRef CAS PubMed.
  26. P. A. Auroux, D. Iossifidis, D. R. Reyes and A. Manz, Anal. Chem., 2002, 74, 2637–2652 CrossRef CAS PubMed.
  27. P. Grodzinski, R. Liu, J. Yang, M. D. Ward and P. Ann, Conf. Proc. IEEE Eng. Med. Biol. Soc., 2004, 26, 2615–2618 Search PubMed.
  28. L. Y. Hung, T. B. Huang, Y. C. Tsai, C. S. Yeh, H. Y. Lei and G. B. Lee, Biomed. Microdevices, 2013, 15, 539–551 CrossRef CAS.
  29. K. Y. Lien, L. Y. Hung, T. B. Huang, Y. C. Tsai, H. Y. Lei and G. B. Lee, Biosens. Bioelectron., 2011, 26, 3900–3907 CrossRef CAS PubMed.
  30. C. H. Tai, Y. C. Tsai, C. H. Wang, T. S. Ho, C. P. Chang and G. B. Lee, Microfluid. Nanofluid., 2014, 16, 501–512 CrossRef CAS.
  31. C. H. Wang, K. Y. Lien, L. Y. Hung, H. Y. Lei and G. B. Lee, Microfluid. Nanofluid., 2012, 13, 113–123 CrossRef CAS.
  32. R. Q. Zhang, S. L. Hong, C. Y. Wen, D. W. Pang and Z. L. Zhang, Biosens. Bioelectron., 2018, 100, 348–354 CrossRef CAS PubMed.
  33. Y. T. Tseng, C. H. Wang, C. P. Chang and G. B. Lee, Biosens. Bioelectron., 2016, 82, 105–111 CrossRef CAS PubMed.
  34. C. H. Wang, C. P. Chang and G. B. Lee, Biosens. Bioelectron., 2016, 86, 247–254 CrossRef CAS PubMed.
  35. Q. Cao, M. Mahalanabis, J. Chang, B. Carey, C. Hsieh, A. Stanley, C. A. Odell, P. Mitchell, J. Feldman, N. R. Pollock and C. M. Klapperich, PLoS One, 2012, 7, e33176 CrossRef CAS PubMed.
  36. Y. Hsu, H. H. Ma, L. S. Lico, J. T. Jan, K. Fukase, Y. Uchinashi, M. M. L. Zulueta and S. C. Hung, Angew. Chem., Int. Ed., 2014, 53, 2413–2416 CrossRef CAS PubMed.
  37. C. H. Weng, K. Y. Lien, S. Y. Yang and G. B. Lee, Microfluid. Nanofluid., 2011, 10, 301–310 CrossRef CAS.
  38. S. Y. Yang, J. L. Lin and G. B. Lee, J. Micromech. Microeng., 2009, 19, 035020 CrossRef.
  39. Y. N. Yang, S. K. Hsiung and G. B. Lee, Microfluid. Nanofluid., 2009, 6, 823–833 CrossRef CAS.
  40. W. H. Chang, S. Y. Yang, C. H. Wang, M. A. Tsai, P. C. Wang, T. Y. Chen, S. C. Chen and G. B. Lee, Sens. Actuators, B, 2013, 180, 96–106 CrossRef CAS.
  41. P. D. Cooper, Adv. Virus Res., 1961, 8, 319–378 CrossRef CAS.
  42. C. Stuart-Harris, Br. Med. Bull., 1979, 35, 3–8 CrossRef CAS.
  43. B. Hoffmann, D. Hoffmann, D. Henritzi, M. Beer and T. C. Harder, Sci. Rep., 2016, 6, 27211 CrossRef CAS PubMed.
  44. K. Tsukamoto, H. Ashizawa, K. Nakanishi, N. Kaji, K. Suzuki, M. Okamatsu, S. Yamaguchi and M. Mase, J. Clin. Microbiol., 2008, 46, 3048–3055 CrossRef CAS PubMed.
  45. M. Abt, J. de Jonge, M. Laue and T. Wolff, Vaccine, 2011, 29, 5153–5162 CrossRef CAS PubMed.
  46. C. Ridenour, A. Johnson, E. Winne, J. Hossain, G. Mateu-Petit, A. Balish, W. Santana, T. Kim, C. Davis, N. J. Cox, J. R. Barr, R. O. Donis, J. Villanueva, T. L. Williams and L. M. Chen, Influenza Other Respir. Viruses, 2015, 9, 263–270 CrossRef CAS PubMed.
  47. L. Y. Hung, J. C. Chang, Y. C. Tsai, C. C. Huang, C. P. Chang, C. S. Yeh and G. B. Lee, Nanomedicine, 2014, 10, 819–829 CrossRef CAS PubMed.
  48. H. C. Lai, C. H. Wang, T. M. Liou and G. B. Lee, Lab Chip, 2014, 14, 2002–2013 RSC.

Footnotes

Preliminary results within this work were presented in the 21st and 22nd International Conference on Miniaturized Systems for Chemistry and Life Sciences (μTAS 2017 and 2018).
Electronic supplementary information (ESI) available. See DOI: 10.1039/c8lc01369a

This journal is © The Royal Society of Chemistry 2019