Thin-film-transistor digital microfluidics for high value in vitro diagnostics at the point of need†

Sally Anderson *, Ben Hadwen and Chris Brown
Sharp Life Science (EU) Ltd, Edmund Halley Road, Oxford Science Park, OX4 4GB, UK. E-mail: sally.anderson2@ntlworld.com; Web: www.aqdrop.com

First published on 30th December 2020

---

---

Introduction

The worldwide impact of COVID-19 has highlighted the need for high quality in vitro diagnostic (IVD) testing for infectious diseases at the point of need (PoN).¹ PoN diagnostics are appealing because they are convenient for the patient, enable timely and targeted treatment, and are important in containing outbreaks of infection.² Microfluidic lab-on-a-chip technologies have long been recognized as being able to make a transformative contribution to this field, in enabling gold-standard IVD testing based on reverse transcription polymerase chain reaction (RT-PCR) to be carried out at PoN.³

aQdrop™ is a droplet digital microfluidics (DMF) lab-on-a-chip in which thin film transistors (TFTs) are used to control droplet movement using the phenomenon of electrowetting-on-dielectric (EWOD).^4,5 aQdrop™ enables multiple droplets to be manipulated simultaneously and independently and was developed by Sharp Life Science (EU) Ltd for liquid handling automation, one example application being nucleic acid library preparation for next generation sequencing (NGS), details of protocols and representative video (S1) can be found in ESI.† In this study we demonstrate how we re-purposed the aQdrop™ platform in response to the COVID-19 pandemic to perform a sample to answer molecular IVD SARS-CoV-2 test. aQdrop™ was used to miniaturise and automate the steps of (i) nucleic acid isolation from saliva spiked with coronavirus RNA, (ii) degradation of double stranded deoxyribonucleic acids (dsDNA) to remove DNA from the sample (iii) analysis of the resultant RNA by reverse transcription quantitative polymerase chain reaction (RT-qPCR) to detect two gene targets: coronavirus RNA (SARS-CoV-2) and the housekeeping gene human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) from the host. The housekeeping gene was included to confirm that sampling had been performed properly and that human cellular material had been analysed.

Approaches to SARS-CoV-2 testing are reviewed by Esbin et al.⁶ Despite multiple alternative testing modalities, RT-PCR remains the gold-standard test type. Reasons for this include; ease of assay design (once the viral genome sequence is known) and high sensitivity due to signal amplification during PCR. Successful RT-PCR testing is, however, dependent on labour intensive procedures carried out in centralised laboratories. Combining microfluidic technology with the detection of viral RNA via RT-PCR is therefore very appealing for a PoN test and several products have received FDA emergency approval or CE mark for COVID-19 diagnosis. These can be grouped into those for detecting only the SARS-CoV-2 pathogen,⁷ and those that test for a panel of respiratory pathogens.⁸

Examples of microfluidic platforms executing RT-PCR tests for a range of target pathogens include QiaStat-Dx (respiratory SARS-CoV-2, Qiagen, Germany),¹ BioFire® FilmArray® (BioFire Diagnostics, LLC, Salt Lake City, UT, USA),⁹ GenMark (Carlsbad, CA, USA, ePlex),^10,11 Bosch (Vivalytic, Bosch Healthcare Solutions, Germany)¹² and Xpert Xpress (SARS-CoV-2/Flu/RSV, Cepheid, CA, USA).¹³ RT-PCR is a two-step reaction in which a target region of the pathogen RNA is first reverse transcribed to complementary deoxyribonucleic acid (cDNA), which is then amplified by PCR or an isothermal amplification method e.g. loop-mediated isothermal amplification (LAMP).¹⁴ The amplified species is then detected, typically by optical means, although for some systems detection is colorimetric or based on electrical methods, e.g. impedance. The testing platforms combine sample preparation and RT-PCR analysis. Sample preparation is critical to prevent inhibitors negatively influencing the nucleic acid amplification process. Although these testing systems are aimed at decentralised settings, they typically work with samples derived from nasopharyngeal swabs, which are taken by healthcare professionals and then eluted into transport media for storage and transport. During analysis the transport media must be transferred to the test cartridge, with potential to contaminate the sample and the instrument operator. The workflow is simplified for the Bosch and Cepheid platforms by inserting swabs directly into the instrument cartridge and breaking off the swab shaft. This process reduces potential contamination but requires sampling to be performed directly before testing. Recent work by Wyllie and co-workers¹⁵ has shown that saliva may be as sensitive for diagnosing COVID-19 as material collected via nasopharyngeal swabs. Saliva is an appealing sample type for PoN IVD tests since sampling may be carried out easily and safely by the patient.¹⁶

Of the PoN tests currently available, those with the largest number of pathogen targets are: (a) the BioFire® FilmArray® System, which automates sample preparation from nasopharyngeal swabs eluted into transport media and pathogen identification via RT-PCR, using a multiplexed panel of 22 viral and bacterial respiratory pathogens including SARS-CoV-2; and (b) the ePlex (RP2 panel; 23 viral and 3 bacterial targets) developed by GenMark and also based on EWOD technology and which provides a sample-to-answer solution for a panel of respiratory pathogens from nasopharyngeal swabs eluted in transport media. The panel approach enables more accurate diagnosis; establishing co-infection and alternative diagnosis, appropriate patient treatment and information for public health in situations where the clinical presentation may be very similar.^17,18 The BioFire® FilmArray® supports qPCR for semi/quantitative target analysis, whereas the ePlex relies on end-point-analysis to provide a ‘yes/no’ answer as to whether a pathogen is present, but gives no information on viral load.

Several challenges remain, and to date PoN infectious disease testing has not yet been widely adopted, for respiratory IVDs in general or SARS-CoV-2 diagnostics in particular. The rate of false negatives is significant,¹⁸ particularly if the sample has low viral load or contains inhibitors. Tests that are based on the detection of a single gene target may be prone to false negatives if the target RNA is damaged or broken in regions corresponding to the PCR primers. For most systems, sufficient data are lacking to give a true measure of test specificity (rate of false positives).⁸ Problems relating to sensitivity and specificity may be compounded by the paucity of available controls. A problem for SARS-CoV-2 testing that extends beyond PoN settings is that the detection of viral RNA may not be correlated with the patient being infected.¹⁹ Correlation of viral detection and infection could in principle be determined by testing for a host response biomarker, though at present such host focused approaches are still in the research stage.^20,21 In conclusion, microfluidic PoN systems do not yet deliver gold-standard molecular IVD tests.

Digital microfluidics (DMF), based on the manipulation of discrete droplets, has emerged as an attractive alternative to traditional channel-based microfluidics.^22–24 DMF devices are comprised of an array of electrodes with an insulating hydrophobic coating. The application of voltages to selective electrodes facilitates lateral movement of droplets by means of the electrowetting effect (EWOD). In traditional “passive” DMF, the number of independently controlled electrodes on the chip is limited to of order 100 by the number of electrical connections that can be made to the consumable microfluidic cartridge. As described earlier, GenMark has commercialised “passive” DMF for PoN IVD. The various steps required for a successful sample-to-answer SARS-CoV-2 test e.g. handling of physiological fluids,²⁵ RNA isolation, RT-PCR²⁶ and qPCR²⁷ have been demonstrated on “passive” EWOD systems.

Thin film transistor digital microfluidics (TFT-DMF, also known as active matrix EWOD and by the commercial name aQdrop™) is a recent development of digital microfluidics,^4,5 which overcomes the limitation in electrode number by implementing control electronics based on TFT technology within the microfluidic cartridge. Very large arrays of independently addressable electrodes are realised, capable of manipulating a large number of droplets of any size and shape, with optimized droplet placement and movement, along any pathway to perform multiple parallel reactions. The provision of a TFT-based droplet sensor capability at each array element ensures very high reliability of droplet operations since these can all be performed with closed-loop feedback control whereby droplets are continually tracked and adjusted. Droplets of different volumes can be metered and measured to accuracies of 2%.⁵ aQdrop is a truly programmable digital microfluidic platform and supports multiple workflows through simple software changes. Examples of automated DNA library preparation on aQdrop can be found in the ESI,† along with an exemplary video.

Combining different assay types (multi-modal testing) on a single testing platform could improve diagnostic accuracy and patient treatment strategies.^28,29 For example, quantitative virus RNA analysis by RT-PCR when combined with immunoassays for antigens and antibodies would potentially reduce the rates of false negative tests. In addition, quantitation of mRNA and protein host response biomarkers may provide valuable prognostic information, such as response to therapy. TFT-DMF has the potential to facilitate this approach – the large format array technology can perform many reactions in parallel, be highly configurable and thus capable of processing a range of sample types e.g. nasopharyngeal swabs or saliva samples, followed by multiple, complex, multi-step assays. These may comprise, for example, a combination of multiplexed RT-qPCR tests for different gene targets, repeats, controls and the combination of molecular diagnostics with assays of other types, e.g. immunoassay detection of host and/or pathogen biomarkers. Finally, leveraging established manufacturing infrastructure, quality control and production techniques from the liquid crystal display industry; high volume, low cost manufacturing compatible with single use disposable tests is possible. We believe that aQdrop™ is well placed to bring gold-standard IVD testing to the PoN.

We report for the first time the automation of the following on aQdrop to generate a sample-to-answer molecular IVD SARS-CoV-2 test to illustrate how the TFT-DMF technology can perform many reactions in parallel and be highly configurable.

1. Bead-based nucleic acid extraction from saliva samples spiked with coronavirus RNA.

2. RT-qPCR of extracted nucleic acid samples to detect and quantify coronavirus RNA spiked into saliva.

3. Extension of the protocol on aQdrop to include degradation of dsDNA after nucleic acid isolation from saliva and qualitative analysis of the resultant RNA sample by RT-PCR (a) for a mRNA target from the host organism (human GAPDH) to confirm successful sample collection (b) to detect the spiked coronavirus RNA.

Methods

1. Thin-film transistor (TFT) chip

The TFT chip design is shown in Fig. 1 (photograph and schematic circuit diagram). In the centre of the chip is an active matrix array of 316 × 130 elements with an element pitch of 210 μm. Driver circuits to control the operation of the array and interface with external instrumentation are arranged at the periphery of the chip. Each array element has an EWOD control electrode of size 207 μm × 207 μm with adjacent electrodes separated by a gap of 3 μm. The control electrodes are formed in indium zinc oxide (IZO) atop the TFT circuitry.

Each array element circuit performs the dual functions of providing an electro-wetting actuation voltage to the electrode and sensing the capacitance presented at each electrode. The design and operation of the array element circuit is a minor modification to that previously reported⁴ and is described in ESI.† The TFT chip was fabricated by Sharp Corporation on a glass substrate (AN100, 0.5 mm thickness, Asahi Glass) using a low temperature polysilicon (LTPS) process which is identical to that used for LCD smartphone display manufacture up to and including the transparent electrode layer. In an additional process step to conventional LCD manufacturing, the TFT chip was coated with a silicon nitride ion barrier insulator layer (300 nm thick, deposited by PECVD).

2. Microfluidic cartridge

For ease of use, and to ensure reliable device operation, the TFT chip was assembled into a single-use microfluidic cartridge. In a first assembly step, the TFT chip and an ITO top glass plate (Asahi Glass, 0.7 mm) were coated with a hydrophobic layer (Cytop, Asahi Glass, 50 nm) and then sandwiched and sealed with a UV cured adhesive to form an EWOD cell with a cell-gap of 240 μm. The combination of the silicon nitride insulator and the CYTOP coating generates a very effective ion barrier. Although the ion barrier is thin, no breakdown was observed with the TFTs used for the development work described in this paper, even when droplets (>500 mM salt) were manipulated under full actuation. A custom designed interface PCB was then attached to the chip connection terminals and the EWOD cell encased within a custom designed black plastic housing. Ports in the plastic housing were aligned with holes in the top plate to enable fluids to be loaded into and extracted from the EWOD cell. The EWOD cell assembly and final microfluidic cartridge construction are shown in Fig. 2. No modifications to the pre-existing microfluidic cartridge were made for the purpose of this work. Further details of the construction are described in ESI.† In operation, a square wave voltage signal of approximately 1 kHz frequency and magnitude up to 20 V is applied to the common ITO electrode on the top plate. To control electrowetting, a voltage signal of similar magnitude and which is either in-phase (non-actuated state) or out-of-phase (actuated state) with the common ITO electrode signal is applied to the control electrode of each element in the array.

3. Control instrument hardware

A control instrument for the microfluidic cartridge was previously developed to support a wide variety of applications on the TFT-DMF platform. The key components of the control instrument are: (a) control electronics – PCBs to generate voltages and timing signals required to drive the cartridge; (b) magnetic system – motor controlled magnets to enable bead-based clean-up steps at eight discrete locations on the active matrix array; (c) thermal system: two motorised Peltier assemblies to control the temperature of droplets within two separate zones on the active matrix array; and (d) optical system – an arrangement of LEDs, filters, lenses and a scientific CMOS camera to enable fluorescence analysis of droplet composition using fluorescein and/or SYBR green fluorophores. An illustration of the arrangement of the microfluidic cartridge with the key internal components of the control instrument is shown in Fig. 3. No modifications to the pre-existing control instrument were made for the purpose of this work. Further details of the design of the control instrument are given in ESI.†

4. Control software

A simplified architecture diagram of the control software is shown in Fig. 4(a). The low-level software modules perform the functions of processing sensor data to determine the size and positions of droplets and define and generate actuation data then communicated to the firmware. The mid-level modules define droplet operations as unitary sequences of actuation data, parameterised by variables such as droplet size and operation execution speed. The droplet operations use feedback, such that an (N + 1)th frame of actuation data is generated from analysis of the Nth frame sensor data. In addition, the mid-level modules also control the thermal and magnetic control systems and the detection optics. High-level modules include the operations scheduler that co-ordinates droplet operations, instrument functions and user interactions and an applications programming interface (API) that defines and exposes the command set of droplet operations and instrument functions. A graphical user interface (GUI) provides the user with control and visualisation and includes a simulation mode to support protocol design. The exemplary GUI screenshot in Fig. 4(b) depicts a single frame of actuation data being transmitted and sensor data being received as multiple droplet operations are in progress. No modifications to the pre-existing control software were made for the purpose of this work.

5. Scripts

Full automation of assays on the TFT-DMF platform is achieved through custom Python scripts. Each automation script defines a complete protocol to implement the desired assay in droplet format on the TFT-DMF device. To perform the protocol, multiple droplet operations including move, split, merge and mix may be defined to occur in series or parallel, chained to perform complex fluidic functions and accurately synchronised as desired. As well as defining the series of droplet operations, instrument functions and user interactions to be executed in a particular protocol, the script also tracks droplets and system status and provides high level error handling and correction.

6. Reagents and kits

7. Analytical tools

Light Cycler 96 (Roche): RT-PCR reactions in the Light Cycler were performed according to the manufacturer specification and the FAM channel was used for readout. Thermal cycling parameters were used as specified for the various master mixes. Tapestation (Agilent) was used with D1000 and genomic tapes to assess by electrophoresis the DNA output of the various steps.

8. Protocol development

Video S2 illustrating operation of aQdrop can be found in the ESI.†

We follow the steps outlined in Fig. 5 during protocol development on aQdrop. The aim of these steps is to successfully transition the chemistry from ‘in tube’ to ‘in droplet’ on aQdrop.

1.1: To check compatibility of kit reagents with the aQdrop genomics filler-fluid, our standard speed and dispense test scripts were run with droplets of all reagents (details of these experiments can be found in the ESI†).

1.2: We used an application specific script, written in a customised version of the Python language, to define a complete protocol to implement the desired protocol steps in droplet format. At this stage we developed a ‘low density script’ where efforts were not made to minimise the protocol execution time through complex parallel droplet operations. The individual steps were carried out on aQdrop and the performance of the steps assessed with standard laboratory analytical equipment as illustrated in Fig. 6.

1.3: The individual steps were combined to provide a sample-to-answer test. Cycle-by-cycle fluorescence images were recorded from RT-PCR and were analyzed using an image processing algorithm implemented in Matlab which determined the boundaries of the droplets from the visible meniscus, determined the average optical signal from the interior of the droplet and subtracted the background signal from an annulus of 2 element extent immediately surrounding the droplet. Outputs from duplicates were averaged before baseline correction and normalization.

Results

aQdrop has been used to automate and miniaturise the steps shown in Fig. 8 to investigate the usefulness of TFT-DMF in bringing gold-standard laboratory tests to PoN for infectious disease diagnostics (in this example SARS-CoV-2) with saliva samples.

Step 1: Nucleic acid extraction with MaxMag kit on aQdrop

A saliva sample in RNA stabilisation buffer spiked with coronavirus RNA (80 copies per μL) or water was used as input to the aQdrop protocol. The number of copies of RNA in the ‘mock’ sample was unrealistically high to make sure that the RNA could be readily detected even in unoptimized workflows.

The layout of droplets for representative reaction steps is illustrated in Fig. 9(A)–(C). User intervention was requested at the start of the protocol to load the genomics filler-fluid and reagents and then at the end of the protocol to extract the nucleic acid Fig. 9(C), otherwise the protocol was performed without user intervention. The TFT chip sensor log was used to check that all protocol steps had been performed as expected.

The resultant purified nucleic acid sample was analyzed off aQdrop on the Tapestation (Fig. 10a) and used as input to RT-PCR in the Light Cycler 96 to confirm the presence of coronavirus RNA. RT-PCR (Primer Design kit) was carried out under standard conditions and as expected coronavirus RNA was detected in samples extracted from saliva spiked with coronavirus RNA, whereas no coronavirus RNA was detected for samples spiked with water. The aQdrop RT-PCR reaction mixtures were also analyzed on the Tapestation (Fig. 10b); those templated with sample derived from coronavirus RNA spiked saliva gave a single band by electrophoresis corresponding to the same amplicon length as samples prepared with positive control DNA in the Light Cycler 96. For the saliva sample spiked with water no bands were observed by electrophoresis.

Step 2: DNAse degradation of dsDNA

Electrophoresis (Tapestation) was used to assess DNAse performance on dsDNA amplicons generated in the Light Cycler 96 during PCR. There was no evidence for any dsDNA after the treatment (see ESI† Fig. S6.5). In addition, when DNAse treatment was incorporated into the sample-to-answer protocol no evidence for genomic DNA was observed in electrophoresis analysis of the droplets after RT-PCR (see ESI† Fig. S6.4).

Step 3a: RT-PCR on aQdrop TFT-DMF

Performance of the aQdrop RT-PCR reactions was found to be similar to that observed for in-tube reactions in the Light Cycler 96, as illustrated in Fig. 11. Thermal cycling and cycle-by-cycle fluorescence imaging resulted in the amplification curves. The solid lines are averages of 4 Light Cycler experiments at 20 copies per μL. Equivalent aQdrop data are shown as data points and are averages from two 1 μL droplets positioned side-by-side and from 4 separate experiments. The difference in turn on between the reactions containing DNA and RNA (same number of copies of nucleic acids) was approximately 2–3 cycles in the Light Cycler and about 3–3.5 cycles for the same reactions carried out on aQdrop. Details of this analysis is summarized in the ESI.†

These experiments demonstrated that DNA positive controls can be misleading when estimating RNA copy numbers, and that RNA positive controls provide a better comparison since they control for both the reverse transcription and the PCR processes.

qPCR – for quantitative DNA analysis on aQdrop

Detection and quantification of cDNA formed during reverse transcription is dependent on reliable aQdrop PCR. To further characterize the thermal cycling performance of aQdrop, RT-qPCR reactions were performed using the DNA standard material from the Primer Design kit as template at a range of concentrations. The cycle-by-cycle amplification curves are illustrated in Fig. 12 and an aQdrop standard curve was generated from these amplification curves and is plotted in Fig. 13. Amplification efficiencies were lower on the TFT-DMF aQdrop platform than in the Light Cycler 96 (see ESI† for standard curves) but still in the range usually considered acceptable.

To determine the sensitivity of aQdrop PCR, the qPCR aQdrop standard curve (performed with 1 μL droplets) was extrapolated to 1 copy per μL, and from this the turn-on cycle number for a single copy of DNA in a 1 μL droplet was estimated to be 31. It was confirmed that turn-on of single copy droplets was detectable with our optical detection system by carrying out droplet digital PCR (Fig. S6.2, ESI†).

Nucleic acid extraction and RT-PCR – one target (steps 1 and 3a)

The aQdrop automated protocols for steps 1 (nucleic acid extraction from saliva) and 3a (RT-qPCR for SARS-CoV-2 detection) were combined into a single script. User intervention was requested both at the start of the protocol to load the genomics filler-fluid and reagents for nucleic acid extraction (Fig. 9(A)–(C)), and then after step 1, to remove waste and to load reagents for analysis of the processed sample by RT-PCR (Fig. 14D). The output from the protocol was a series of cycle-by-cycle fluorescence images from RT-qPCR, and a log file of the TFT chip sensor output detailing the progress of every droplet on aQdrop. The sensor output illustrating RT-qPCR reaction set-up and droplet positions for thermal cycling is illustrated in Fig. 14E and RT-qPCR amplification curves derived from the fluorescence images in Fig. 15. A time lapse video (Video S3†) showing the cycle-by-cycle changes in droplet fluorescence can be found in the ESI.† The final image of the thermally cycled droplets is inset on the RHS in Fig. 15; the positive controls (C1 and D1) and sample droplets (A1) are strongly fluorescent while the fluorescence output from the negative control droplets (B1) is unchanged by thermal cycling.

Nucleic acid extraction and RT-PCR – two targets (steps 1, 2 and 3)

Steps to degrade dsDNA and qualitatively analyze for mRNA from the host organism were added to the protocol (all steps in Fig. 8). As for the previous example, user intervention was requested at the start of the protocol and then at the end of step 1 to remove waste and add the reagents for dsDNA degradation and RT-PCR. As before, the output from the automated protocol was a series of cycle-by-cycle fluorescence images and a log file detailing the progress of every droplet on aQdrop. RT-PCR amplification curves were derived from the fluorescence images (Fig. 17). The sensor output (Fig. 16) illustrates how the output from the nucleic acid extraction (step 1, Fig. 8) was merged, mixed and incubated with DNAse (Fig. 16D) before being split for spatially multiplexed RT-PCR for coronavirus and human GAPDH targets (Fig. 16E). A video (Video S4†) of the TFT chip sensor output for the complete protocol can be found in the ESI.

Exemplary optical detection with a smartphone

An important requirement for PoN is that the instrument should be small and relatively low cost. To demonstrate the feasibility of optical detection with a miniaturised and low-cost camera module. The scientific CMOS camera and lens was replaced by a smartphone camera (iPhone 11 Pro, standard lens) that was remotely triggered by the instrument control electronics. Results are shown in Fig. S6.9, ESI.†

Discussion

We demonstrated ‘yes/no’ assessment of a saliva sample for coronavirus RNA, showed that the sample was from a human patient (detection of human GAPDH) and confirmed that the PCR process was effective on aQdrop. We also showed that with inclusion of both DNA and RNA controls, performance of both RT and PCR steps may be assessed independently and a good estimate of the virus RNA copy number in the original sample calculated. By extrapolation, similar improvements could be made for housekeeping genes (e.g. human GAPDH) enabling assessment of the number of host cells in the original sample and confidence that the sample had been correctly collected and stored before testing.

Saliva is a challenging sample type to manipulate by electrowetting²⁵ and for it to be useful for gene expression analysis³⁰ it must be processed carefully to remove proteins and microbial nucleic acids. We have shown that nucleic acid extraction from saliva may be performed on aQdrop with standard laboratory kit reagents, and that subsequent removal of dsDNA is possible with DNAse treatment. This is an exemplary set of experiments and future studies would look to test various RNA extraction and purification strategies to maximise the diagnostic utility of aQdrop. In addition, there is also a need to process more sample, particularly when the number of virus particles is low, as is the case at the start of infection, and to introduce a method of sample concentration before loading on aQdrop. Rackus et al.³¹ described a paper-based method to concentrate analytes bound to magnetic beads and Jebrail et al. a companion module to process large volumes.³²

RT-qPCR was found to perform as expected in-droplet and mirrored that for in-tube reactions carried out using a conventional laboratory thermal cycler (Light Cycler 96). Differences between different enzyme systems were observed and were attributed to efficiency of the reverse transcription enzymes in the 1-step master mixes.³³ This work employed 1-step RT-qPCR, combining the chemistry for reverse transcription and amplification in a single droplet. The ability to manipulate droplets discretely gives the highly desirable possibility of separating the RT and qPCR steps (2-step RT-qPCR).³⁴ This affords more flexibility in the choice of RT enzymes and DNA polymerases and is the preferred choice for low quantities of starting material.³⁵ For example, a single clean-up and RT step could be used to create a pool of cDNA, from which samples are split off and combined with primers and probes suitable for analysis for a range of targets. This may be described as spatial multiplexing, with different reactions at different regions of the chip. With appropriate fluorescence-based detection optics as part of the instrument, optical multiplexing is also possible, using different wavelength probes to detect multiple targets in the same droplet.

Although our initial studies have been limited to assessing samples for two RNA targets, it is easy to see that with the parallelism afforded by the TFT-DMF platform to manipulate many droplets independently and simultaneously, further RNA targets could also be quantified to assess the host's response to the virus, e.g. detecting and characterising mRNAs that result from changes in biological processes triggered in response to the infecting pathogen,²¹ to widen the number of targets on the viral genome of interest and to include targets for other common viral respiratory diseases. In the case of COVID-19, having the ability to determine viral load has clinical relevance.³⁶ In addition, since TFT-DMF has already been shown to support immunoassays (to test for antigen proteins and antibodies), different assay types could also be carried out in parallel for a multi-modal diagnostic testing approach at PoN.²⁸

In these proof of concept assessments, the focus has been on confirming the chemistry performance and not on maximising reaction density (see ESI† for representative reaction densities for DNA library preparation protocols) and number of parallel processes or on minimising hands-on time. Hence, aqueous waste was removed from the device and reagents for RT-qPCR added after the MagMax™ nucleic acid extraction protocol (step 1) had completed. Our intention is to remove this manual intervention, and for all reagents to be loaded at the start of a protocol.

A video showing the cycle-by-cycle increase in fluorescence can be found in the ESI.† The negative control droplets processed in the same region of the aQdrop device as the RNA containing saliva samples showed no increase in fluorescence during RT-PCR suggesting that droplet-to-surface-to-droplet contamination is not significant.

The on-chip RT-qPCR amplification curves are noisier than those derived from Light Cycler amplification, a function of the miniaturisation of the assay such that the optical signals are small. The aQdrop device also generates a background fluorescent signal, a consequence of the organic planarization layers that are part of the LTPS process. It has separately been shown that the chip background fluorescence can be reduced by 1–2 orders of magnitude by fabricating the TFT with an opaque (reflective) metal (e.g. molybdenum) electrode instead of transparent IZO (opaque metal electrode material is available as a standard variant, being used to manufacture reflective type LC displays).

Conclusions and outlook

This work has demonstrated the latest development in TFT-DMF microfluidic cartridges and shown potential application to molecular diagnostics (e.g. for SARS-CoV-19) at the point-of-need. A state of the art TFT backplane was described, featuring input output multiplexing, and selective row addressing to realise a 41k array element chip capable of high frame rate addressing and sensor readout.

We further demonstrated an aQdrop SARS-CoV-2 RT-qPCR detection sample-to answer protocol featuring automated sample extraction and lysis, bead-based clean-up, dsDNA degradation, reverse transcription, and real time PCR amplification with fluorescence detection.

We discuss the potential of TFT-DMF technology to meet the need for PoN testing for respiratory infectious disease, featuring gold-standard quality of result, capability to perform a panel of tests including those to guide antibiotic use,³⁶ and the possibility to perform multi-modal tests that explore the host response to infection, e.g. molecular diagnostics, in parallel with immunoassays, all at low cost and with rapid scale-up potential.

Future development will be needed to realise a PoN TFT-DMF microfluidic cartridge including development of capability to store reagents within the cartridge and to provide a foolproof and easy to use method of introducing the sample.

Creation of a low cost, miniaturised instrument is also thought to be feasible; this work (and others, for example Farshidfar et al.³⁷) have demonstrated the capability to perform optical detection with a smartphone camera, and miniaturised thermal cycling capabilities are also well known.³⁸ Miniaturisation of the other instrument components (magnetic system, electronics) is not expected to impose any particular technical difficulties.

The features demonstrated in this paper lay the foundations important to achieving laboratory quality diagnostic results at PoN.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work would not have been possible without many former colleagues at Sharp Life Science (EU) Ltd. who have contributed to the development of the aQdrop™ Thin Film Transistor Digital Microfluidics platform. We would particularly like to acknowledge: Emma Walton, Lesley Parry Jones, Sinéad Matthews and Christine Hawkins who contributed to the design and development of the microfluidic cartridge, Oliver Beard who co-designed the TFT-DMF electronics backplane, the SLS software team, in particular Rob Amos, Tim Boyle, Rob Pym, Umesh Keswani and Ome Peddi who designed and created the control software. Helen Forrest and Aengus Rae who developed original Python scripts upon which the scripts of this work were based and Simon Bryant who performed early experiment on bead-based DNA isolation methods. Gregory Gay and Alexander Tookey who developed the methods for removal of trapped oil from the droplet interior. Gregory Gay for making the coloured dye DNA NGS library preparation video. Jonathan Buse who designed the instrument control electronics and wrote the firmware. Phil Roberts who designed the instrument magnetic and thermal control hardware, the SLS production team (Alan Green, Jason Bibby and Bruce Culbert) who modularised the cartridges into plastic housings and assembled the instrument. Andrew Kay (of Sharp Laboratories of Europe Ltd) who created the Matlab routine for analysing the optical images. Ishna Mallinson, Jordan Brown, Claire Ferrao, and Rebecca Sanders who carried out protocol development for DNA library preparation on aQdrop. Adam Wilson for writing the R script to analyse speed and dispense test data. Adrian Jacobs, Campbell Brown, Adam Nightingale and Pamela Dothie who made many contributions to the development of the platform over the years. The authors would like to thank Akira Imai and Takeshi Hara and their colleagues at Sharp Corporation for fabricating the TFT chips and EWOD cells.

Notes and references

1. N. J. Brendish, S. Poole, V. V. Naidu, C. T. Mansbridge, N. J. Norton, H. Wheeler, L. Presland, S. Kidd, N. J. Cortes, F. Borca, H. Phan, G. Babbage, B. Visseaux, S. Ewings and T. W. Clark, Lancet Respir. Med., 2020, 8, 1192–1200,  DOI:10.1016/S2213-2600(20)30454-9 .
2. J. Dinnes, J. J. Deeks, A. Adriano, S. Berhane, C. Davenport, S. Dittrich, D. Emperador, Y. Takwoingi, J. Cunningham, S. Beese, J. Dretzke, L. Ferrante di Ruffano, I. M. Harris, M. J. Price, S. Taylor-Phillips, L. Hooft, M. M. G. Leeflang, R. Spijker and A. Van den Bruel, Cochrane Database of Systematic Reviews, 2020,(8), CD013705,  DOI:10.1002/14651858.CD013705 .
3. S. Park, Y. Zhang, S. Lin, T.-H. Wang and S. Yang, Biotechnol. Adv., 2011, 29, 830–839 CrossRef CAS .
4. B. Hadwen, G. R. Broder, D. Morganti, A. Jacobs, C. Brown, J. R. Hector, Y. Kubota and H. Morgan, Lab Chip, 2012, 12, 3305–3313 RSC .
5. S. Kalsi, M. Valiadi, M.-N. Tsaloglou, L. Parry-Jones, A. Jacobs, R. Watson, C. Turner, R. Amos, B. Hadwen, J. Buse, C. Brown, M. Sutton and H. Morgan, Lab Chip, 2015, 15, 3065–3075 RSC .
6. M. N. Esbin, O. N. Whitney, S. Chong, A. Maurer, X. Darzacq and R. Tjian, RNA, 2020 DOI:10.1261/rna.076232.120 .
7. M. M. Gibani, C. Toumazou, M. Sohbati, R. Sahoo, M. Karvela, T.-K. Hon, S. De Mateo, A. Burdett, K. Y. F. Leung, J. Barnett, A. Orbeladze, S. Luan, S. Pournias, J. Sun, B. Flower, J. Bedzo-Nutakor, M. Amran, R. Quinlan, K. Skolimowska, C. Herrera, A. Rowan, A. Badhan, R. Klaber, G. Davies, D. Muir, P. Randell, D. Crook, G. P. Taylor, W. Barclay, N. Mughal, L. S. P. Moore, K. Jeffery and G. S. Cooke, Lancet Microbe, 2020, 1(7), e300–e307,  DOI:10.1016/S2666-5247(20)30121-X .
8. N. Ravi, D. L. Cortade, E. Ng and S. X. Wang, Biosens. Bioelectron., 2020, 165, 112454 CrossRef CAS .
9. H. M. Creager, B. Cabrera, A. Schnaubelt, J. L. Cox, A. M. Cushman-Vokoun, S. M. Shakir, K. D. Tardif, M.-L. Huang, K. R. Jerome, A. L. Greninger, D. Drobyshevae, U. Spaulding, M. Rogatcheva, K. M. Bourzac, S. H. Hinrichs, M. J. Broadhurst and P. D. Fey, J. Clin. Virol., 2020, 129, 104538 CrossRef CAS .
10. GenMark RP2 panel (viewed 01/11/2020), https://www.genmarkdx.com/wp-content/uploads/2020/07/PI1114B-ePlex-RP2-PI-EUA-1.pdf .
11. R. H. T. Nijhuis, D. Guerendiain, E. C. J. Claas and K. E. Templeton, J. Clin. Microbiol., 2017, 55, 1938–1945 CrossRef CAS .
12. Bosch multiplex panel including SARS-CoV-2 (viewed 01/11/2020), https://www.bosch-vivalytic.com/en/product/vivalytic-tests/vri-multiplex-test/ .
13. Cepheid Xpert® Xpress SARS-CoV-2/Flu/RSV (viewed 01/11/2020), https://www.cepheid.com/coronavirus-product-resources .
14. M. J. Kellner, J. J. Ross, J. Schnabl, M. P. S. Dekens, R. Heinen, I. Grishkovskaya, B. Bauer, J. Stadlmann, L. Menéndez-Arias, R. Fritsche-Polanz, M. Traugott, T. Seitz, A. Zoufaly, M. Födinger, C. Wenisch, J. Zuber, Vienna COV-19 Diagnostics Initiative (VCDI), A. Pauli and J. Brennecke, bioRxiv, 2020, preprint  DOI:10.1101/2020.06.23.166397.
15. A. L. Wyllie, J. Fournier, A. Casanovas-Massana, M. Campbell, M. Tokuyama, P. Vijayakumar, J. L. Warren, B. Geng, M. C. Muenker, A. J. Moore, C. B. F. Vogels, M. E. Petrone, I. M. Ott, P. Lu, A. Venkataraman, A. Lu-Culligan, J. Klein, R. Earnest, M. Simonov, R. Datta, R. Handoko, N. Naushad, L. R. Sewanan, J. Valdez, E. B. White, S. Lapidus, C. C. Kalinich, X. Jiang, D. J. Kim, E. Kudo, M. Linehan, T. Mao, M. Moriyama, J. E. Oh, A. Park, J. Silva, E. Song, T. Takahashi, M. Taura, O.-E. Weizman, P. Wong and S. Bermejo, N. Engl. J. Med., 2020, 383, 1283–1286 CrossRef .
16. I. M. Ott, M. S. Strine, A. E. Watkins, M. Boot, C. C. Kalinich, A. Harden, C. B. F. Vogels, A. Casanovas-Massana, A. J. Moore, M. C. Muenker, M. Nakahata, M. Tokuyama, A. Nelson, J. Fournier, S. Bermejo, M. Campbell, R. Datta, C. S. Dela Cruz, S. F. Farhadian, A. I. Ko, A. Iwasaki, N. D. Grubaugh, C. B. Wilen and A. L. Wyllie, medRxiv, preprint, 2020,  DOI:10.1101/2020.08.03.20165233.
17. N. J. Brendish, A. K. Malachira, K. R. Beard, S. Ewings and T. W. Clark, Eur. Respir. J., 2018, 1800555 CrossRef .
18. S. L. Mitchell and K. St. George, J. Clin. Virol., 2020, 128, 104429 CrossRef CAS .
19. T. Jefferson, C. Heneghan, E. Spencer and J. Brassey, Are you infectious if you have a positive PCR test result for COVID-19?, https://www.cebm.net/covid-19/infectious-positive-pcr-test-result-covid-19/ .
20. T. W. Clark, N. J. Brendish, S. Poole, V. V. Naidu, C. Mansbridge, N. Norton, H. Wheeler, L. Presland and S. Ewings, J. Infect., 2020, 81, 607–613 CrossRef CAS .
21. S. A. Thair, Y. D. He, Y. Hasin-Brumshtein, S. Sakaram, R. Pandya, J. Toh, D. Rawling, M. Remmel, S. Coyle, G. N. Dalekos, I. Koutsodimitropoulos, G. Vlachogianni, E. Gkeka, E. Karakike, G. Damoraki, N. Antonakos, P. Khatri, E. J. Giamarellos-Bourboulis and T. E. Sweeney, medRxiv, 2020, preprint,  DOI:10.1101/2020.06.18.20131326.
22. R. B. Fair, Microfluid. Nanofluid., 2007, 3, 245–281 CrossRef CAS .
23. E. Samiei, M. Tabrizian and M. Hoorfar, Lab Chip, 2016, 16, 2376–2396 RSC .
24. J. Li and C.-J. Kim, Lab Chip, 2020, 20, 1705–1712 RSC .
25. V. Srinivasan, V. K. Pamula and R. B. Fair, Lab Chip, 2004, 4, 310–315 RSC .
26. A. Rival, D. Jary, C. Delattre, Y. Fouillet, G. Castellan, A. Bellemin-Comte and X. Gidrol, Lab Chip, 2014, 14, 3739–3749 RSC .
27. Z. Hua, J. L. Rouse, A. E. Eckhardt, V. Srinivasan, V. K. Pamula, W. A. Schell, J. L. Benton, T. G. Mitchell and M. G. Pollack, Anal. Chem., 2010, 82, 2310–2316 CrossRef CAS .
28. R. S. Sista, R. Ng, M. Michael, J. Coyne, J. Elderbroom, D. Hull, K. Kay, M. Krishnamurthy, C. Roberts, D. Wu, A. D. Kennedy, R. Singh, V. Srinivasan and V. K. Pamula, Diagnostics, 2020, 10, 10010021 CrossRef .
29. K. Green, S. Graziadio, P. Turner, T. Fanshawe and J. Allen, https://www.cebm.net/covid-19/molecular-and-antibody-point-of-care-tests-to-support-the-screening-diagnosis-and-monitoring-of-covid-19/ .
30. P. Ostheim, A. Tichy, I. Sirak, M. Davidkova, M. Majewski, M. Port and M. Abend, Sci. Rep., 2020, 10, 11147 CrossRef CAS .
31. D. G. Rackus, R. P. S. de Campos, C. Chan, M. M. Karcz, B. Seale, T. Narahari, C. Dixon, M. Dean Chamberlain and A. R. Wheeler, Lab Chip, 2017, 17, 2272–2280 RSC .
32. M. J. Jebrail, A. Sinha, S. Vellucci, R. F. Renzi, C. Ambriz, C. Gondhalekar, J. S. Shoeniger, K. D. Patel and S. S. Branda, Anal. Chem., 2014, 86, 3856–3862 CrossRef CAS .
33. N. Minshall and A. Git, Sci. Rep., 2020, 10, 8151 CrossRef CAS .
34. Thermo Fisher Scientific Website, What's the difference between one-step and two-step RT-PCR? (viewed 01/11/2020), https://www.thermofisher.com/uk/en/home/brands/invitrogen/molecular-biology-technologies/spotlight-articles/onestep-vs-twostep-rtpcr.html .
35. E. Pujadas, F. Chaudhry, R. McBride, F. Richter, S. Zhao, A. Wajnberg, G. Nadkarni, B. S. Glicksberg, J. Houldsworth and C. Cordon-Cardo, Lancet Respir. Med., 2020, 8(9), e70,  DOI:10.1016/S2213-2600(20)30354-4 .
36. Z. Dhesi, V. I. Enne, J. O‘Grady, V. Gant and D. M. Livermore, ACS Pharmacol. Transl. Sci., 2020, 3, 401–417 CrossRef CAS .
37. N. Farshidfar and S. Hamedani, Mol. Diagn. Ther., 2020, 24, 371–373 CrossRef CAS .
38. T. G. Montague, A. Almansoori, E. J. Gleason, D. S. Copeland, K. Foley, S. Kraves and E. Alvarez Saavedra, PLoS One, 2018, 13, e0205852 CrossRef .

---

Footnote

† Electronic supplementary information (ESI) available: Videos S1–S4. See DOI: 10.1039/d0lc01143f

This journal is © The Royal Society of Chemistry 2021