Vibration mixing for enhanced paper-based recombinase polymerase amplification

Abstract

Isothermal nucleic acid amplification tests (NAATs) are a vital tool for point-of-care (POC) diagnostics. These assays are well-suited for rapid, low-cost POC diagnostics for infectious diseases compared to traditional PCR tests conducted in central laboratories. There has been significant development of POC NAATs using paper-based diagnostic devices because they provide an affordable, user-friendly, and easy to store format; however, the difficulties in integrating separate liquid components, resuspending dried reagents, and achieving a low limit of detection hinder their use in commercial applications. Several studies report low assay efficiencies, poor detection output, and poorer limits of detection in porous membranes compared to traditional tube-based protocols. Recombinase polymerase amplification is a rapid, isothermal NAAT that is highly suited for POC applications, but requires viscous reaction conditions that has poor performance when amplifying in a porous paper membrane. In this work, we show that we can dramatically improve the performance of membrane-based recombinase polymerase amplification (RPA) of HIV-1 DNA and viral RNA by employing a coin cell-based vibration mixing platform. We achieve a limit of detection of 12 copies of DNA per reaction, nearly 50% reduction in time to threshold (from ~10 minutes to ~5 minutes), and an overall fluorescence output increase up to 16-fold when compared to unmixed experiments. This active mixing strategy enables reactions where the target and reaction cofactors are isolated from each other prior to the reaction. We also demonstrate amplification using a low-cost vibration motor for both temperature control and mixing, without the requirement of any additional heating components.

Graphical Abstract

Isothermal nucleic acid amplification tests are a vital tool for point-of-care diagnostics. In this work, we significantly improve paper-based recombinase polymerase amplification of HIV-1 DNA and RNA by employing a vibration-based mixing platform.

Introduction

Nucleic acid amplification tests (NAATs) are used to detect infectious diseases due to their high clinical sensitivity and specificity. PCR is the gold standard NAAT that amplifies pathogenic nucleic acids using a thermocycling process to denature, anneal, and extend target.¹ PCR typically operates in central laboratories where expensive automated platforms conduct high throughput testing due to the requirement of thermocycling, high cost instrumentation, cold chain logistics for reagents, trained staff, and infrastructure to store required equipment.² In outpatient clinics, this central lab-based diagnostic testing model can result in delayed diagnoses that prevent immediate follow up treatment.³ Consequently, there is a need for affordable point-of-care (POC) NAATs suitable for outpatient, resource-limited clinics, especially in low- or middle-income countries where disease burden is high.²

The application of PCR for low-cost NAATs is challenging due to the complexity and cost associated with the precise and rapid temperature control required for thermal cycling. The runtime of PCR tests is relatively slow, typically ranging from 45 - 90 minutes. Isothermal amplification methods are a solution to the limitations of PCR for POC diagnostic devices. Isothermal NAATs can match the sensitivity and specificity of PCR, while operating at a single incubation temperature, alleviating the need for rapid and precise temperature control hardware. Isothermal methods, such as loop-mediated isothermal amplification (LAMP), are gaining research interest as they can be performed at a single temperature and detected using a variety of visual or fluorescent detection methods. Some difficulties in the application of LAMP for POC NAATs include the relatively long amplification times of 30 - 60 minutes and the requirement of four to six primers for the assay, which results in complex primer design and high potential for false positives due to nonspecific primer dimer interactions.^4,5 Recombinase polymerase amplification (RPA) is an attractive isothermal amplification method as it produces results in less than 15 minutes, features a low incubation temperature (37 - 40 °C), and has high specificity due to its two primers and sequence-specific probe for fluorescence or lateral flow readout.^6,7

Microfluidic paper-based analytical devices (μPADs) are gaining popularity due to their low cost and useful characteristics for the application of POC NAATs. Many POC NAATs use the term “paper-based” to refer to the use of cellulose, nitrocellulose, polyethersulfone, glass fiber, polycarbonate, and other types of fibers or membranes.^8-15 For this reason, we will henceforth use the term paper-based to refer to NAATs that implement any of these membranes, fibers, or filters. Paper is conducive to stack, store, transport, and incorporate into high volume manufacturing lines. Chemical and biological reagents can be dried or lyophilized on membrane, similar to lateral flow rapid tests.^16-18 When dry, the wicking properties of paper facilitates the transport of solutions. μPAD diagnostic devices often leverage paper’s wicking properties by layering or folding membranes over each other to increase contact area between liquids, transfer target between separate assay steps, or contain specific reaction volumes while removing excess buffer.^19,20 Many labs implemented POC μPADs for diagnostics using these sliding, layering, or folding techniques in their devices.^15,20-22 While paper POC devices are being used for complex assays, they often struggle with uneven analyte concentration profiles after reagent rehydration and assay integration steps.^19,23-25 Furthermore, the pore structure and surface properties of analytical membranes may inhibit nucleic acid amplification reactions through adsorption and/or limiting the diffusion of necessary biomolecular reagents, such as nucleic acids, oligonucleotides, and enzymes needed for amplification reactions. Several studies reported low assay efficiencies, poor detection output, and poorer limits of detection in porous membranes compared to traditional tube-based protocols.^9,15,26

Limited diffusion in paper-based RPA assays are specifically challenging because of the polymer crowding agent that is critical to stabilize enzymes and improve reaction efficiency.^27,28 The addition of crowding agents (typically polyethylene glycol) significantly increases viscosity and reduces diffusion of reactants in the system. For tube-based assays, the high viscosity of RPA buffers requires an intermediate mixing step at 4-5 minutes into the incubation to redistribute RPA reactants and improve amplification output signal.²⁸ Improved RPA reaction efficiency has also been observed under continuously stirred conditions.²⁹ Mixing is difficult to achieve in membrane-based RPA, and may have hampered the development of paper-based RPA. Several paper-based RPA assays have dealt with the mixing challenges in different ways. Sullivan et al. leveraged the high viscosity RPA buffer and diffusion limiting membrane for quantitative detection of RPA; however, the overall fluorescence product was low and the reported limit of detection (LOD) was higher than tube-based assays.⁹ Other paper-based RPA assays showed that assays run in membranes have worse LODs and signal to noise ratios than tube-based ones.^8,15 Overall, the paper-based RPA studies in the literature primarily address integration of reagents and target before amplification but have found poor to moderate LODs and low overall output signal. Improving mixing of paper-based RPA reactions may overcome issues with low diffusion and associated inhibition of RPA, enabling the use of low cost optical or electronic detection strategies and improving LOD.

Mixing, typically defined as bulk stirring with molecular diffusion,³⁰ is difficult to accomplish in membrane-based RPA reactions due to the high viscosity buffer, the porous structure of membranes, the low Reynolds number surface tension driven flow, and the requirement of vigorous mixing for high reaction efficiencies. Most examples of microfluidic mixing in porous substrates use passive mixing methods, which rely on diffusion by increasing contact area between separate solutions.^31,32 The stringent requirements for RPA makes it difficult to achieve efficient mixing using passive mixing methods. Active mixing methods addresses these problems as it uses external driving forces to produce efficient mixing and can be controlled with its power input.^19,24,33-35 The limited examples of active membrane mixing include Y-shaped paper channels with surface acoustic waves or piezoelectric transducers.^36,37 Active mixing methods can be difficult to integrate into paper-based POC devices as complex lithographic fabrication is typically required and voltage requirements are considerably high (PZTs require 30-120 volts).^33,34,36,37 A simple active mixing strategy such as vibration based mixing was demonstrated to be effective in microfluidic devices and large liquid volumes.^38-41 They can use simple cost-effective solutions such as loudspeaker tweeters or eccentric rotating mass (ERM) motors. Motors such as coin cell ERM motors are typically found in cellphones or tablets and provide an attractive solution as they are robust, have a compact profile, are low cost, and require little voltage (0-5 volts). Microfluidic POC devices that have implemented these motors use them for lysis or sample prep applications in liquid volumes.^42-46 While there are many examples of mixing strategies deployed in microfluidic chips, we are not aware of any studies that have explored simple active mixing in paper-based membranes for the enhancement of paper-based NAATs.

In this paper, we show that vibrational mixing from an ERM coin cell motor in paper-based fluidic systems improves mixing, RPA reaction efficiency, overall product output, and reduces time to detection. We use HIV-1 as the diagnostic target in this work since there is a prevailing need for highly-sensitive POC NAATs for HIV viral load monitoring (target LOD <1,000 cp/mL).^47-49 We show that vibrational mixing of paper-based RPA enables the detection of 12 copies of HIV proviral DNA per reaction and 50 copies of HIV RNA per reaction. Mixing in paper-based reactions improves the LOD by a factor of 6.4, reduces detection time from an average of 9.93 minutes to 5.17 minutes, and produces an overall fluorescence signal up to 15.8 times brighter compared to reactions without mixing. Our mixing device achieves the optimal 38 °C incubation temperature of RPA using a self-regulating positive temperature coefficient (PTC) heater. We demonstrate the heat generation from the coin vibration motor may be used on its own without a PTC for heating. We demonstrate that nonuniform initial distributions of reactants drastically inhibit unmixed paper-based RPA and that vibrational mixing readily overcomes these concentration gradients, enabling robust detection at low copy numbers. This platform addresses challenges facing membrane-based POC diagnostic devices such as uneven distribution of reagents, integration of separated components or reagents, and mixing of low volume and viscous samples. Our mixing strategy for improved paper-based diagnostics is inexpensive, simple to implement, and has low power consumption. This mixing platform can also be applied to other membrane-based assays or processes that require high reaction efficiency, rehydration of reagents, high signal output, and device integration with multiple separated steps. Similar isothermal amplification methods such as membrane LAMP assays, sample preparation, or enzymatic reactions may also benefit from efficient membrane-based mixing.

MATERIALS AND METHODS

I. Device Fabrication

The vibrational mixing device consists of a 3D printed assembly that integrates the motor and PTC heater with a glass slide that supports the amplification membrane, as shown in Figure 1. The 3D printed part consistently positions the RPA membrane on the glass slide relative to the motor for each experiment, ensuring repeatable results. The motor assembly consists of a vibration coin cell motor (VC1234B016F, Vybronics, CN), a PTC heater (PTC Heating Element, DFRobot, CN), an isolation material, and a 3D printed plate (S Series Tough PLA, Ultimaker, USA). We use epoxy adhesive (BONDiT B-45, Reltek, USA) and thin double-sided adhesive (CBGNHW011091, Scotch, USA) to attach the foam and coin cell motor to the 3D printed base plate. Both motor and glass slide holder components attach to a base acrylic plate (8505K748, McMaster-Carr, USA) cut on a CO₂ laser (PLS6.150D, Universal Laser Systems, USA). We use commercially available isolation materials in varying thicknesses ranging from 4.75 - 8 mm (DualPlex 6 mm Neoprene, Honwally 6 mm Silicone Gel, Thermwell 6 mm Polyfoam, Thermwell 4.75 mm PVC based foam, and Thermwell 8 mm rubber foam) between the base plate and PTC heater to assess the effect of isolation materials on mixing and resultant RPA performance. We chose low-cost, off the shelf materials that were close in thickness while providing a sufficient variety in material type. Here we focus on screening off the shelf materials for the best mixing performance rather than a systematic and rigorous evaluation of material composition, thickness, and its impact on mixing. To account for height differences, we adjusted the 3D printed platform for each isolation material to maintain the same distance between the motor and glass fiber reaction membrane. Glass fiber membranes (GF/DVA, Whatman, UK) are cut on a flatbed cutter (FCX4000-50ES, Graphtec America, Inc., USA) into uniform squares with a 7.2 mm side length and 785 μm thickness to fit a 50 μL reaction for RPA and RT-RPA assays. We place glass fiber amplification membranes onto a 2” X 3” glass slide (CA6101, Premiere, CN). After adding 50 μL to the membrane, we fully seal the reaction with PCR film (TempPlate RT Select Optical Film, USA Scientific, USA) and slot it into the device with the PCR film in contact with the motor. The coin vibration motor uses a 2 volt input, unless otherwise specified.

Figure 1.

(A) A simplified representation of homogenized membrane-based RPA reactions for unmixed and mixed scenarios. When continuously mixed, reaction products and RPA reagents are redistributed, resulting in increased amplification regions and an overall improvement in reaction efficiency. (B) Image of platform cross section with RPA reaction plate, which consists of: a glass slide, RPA membrane, and PCR film. (C) Photo of platform. The white, dashed-line box represents the location of the RPA membrane in the experiment. The x and y axis are in plane to the motor and the z-axis is perpendicular to the motor face.

II. Visualization of bulk stirring

To visualize and validate bulk stirring due to the vibrational motor, we track fluorescently labeled DNA over 20 minutes of mixing in a liquid saturated membrane. To maintain the viscous and proteinaceous reaction conditions of RPA, we rehydrate a lyophilized RPA TwistAmp exo pellet with 29.5 μL of the TwistAmp reaction buffer and 16 μL of water. We pipet 4.5 μL of 70 base pair DNA tagged with Alexa Fluor 488, taken from the pol region of the HIV-1 genome, onto the center of the membrane. For a homogenized control experiment, fluorophore-tagged DNA is initially and uniformly distributed within the RPA reaction mixture (via vortexing) before pipetted evenly over the membrane.

III. RPA and RT-RPA Reactions

For RPA experiments, we target synthetic DNA (gBlocks Gene Fragments, Integrated DNA Technologies, USA) that contain 1000 base pairs of the HIV-1 genome (group M, subtype A). RT-RPA amplifies purified HIV RNA using HIV-1 supernatant, as described in Lillis et al.⁷ HIV-1 supernatant (group M, subtype A, NCBI accession number: JX140650) provided by the External Quality Assurance Program Oversight Laboratory at Duke University, and is extracted using QIAamp Viral RNA Mini Kits (Qiagen, DEU).⁵⁰ The viral RNA concentration is verified with quantitative real-time PCR. Both targets are diluted in DEPC-treated water and are used in establishing the quantifiable ranges for RPA amplification. The primers and probe used in both RPA and RT-RPA are developed by Lillis et al. for cross-subtype HIV-1 detection.⁷ The RPA mastermix consists of a TwistAmp exo kit lyophilized pellet (TwistDx, UK), 29.5 μL rehydration buffer, 14 mM magnesium acetate, 540 nM forward and reverse primer (Integrated DNA Technologies, USA), and 120 nM exo-probe (LGC Biosearch Technologies, GBR). For RT-RPA experiments, 0.5 U μL⁻¹ of reverse transcriptase (Agilent AffinityScript, USA) is added to the mastermix. The mastermix consisting of rehydration buffer, primers, probe, and reverse transcriptase (for RT-RPA experiments) is added to a lyophilized exo-kit RPA pellet.

IV. Paper-based RPA reactions

We conduct homogenized paper-based RPA experiments with mastermix, magnesium, and target uniformly distributed onto a reaction membrane. For this set of homogenized experiments, target (DNA or RNA) and MgOAc are added to the cap of the RPA tube. We close the tube and shake it manually for 20-30 seconds to ensure homogeneous distribution of reactants. Then, we evenly pipet 50 μL of the RPA reaction (mastermix, magnesium and target) onto the membrane before sealing the amplification membrane with PCR film and slotting it into the assembly. We set the voltage on the PTC heater voltage to result in 37-39 °C output on the face of the motor to provide heat to the RPA reaction. Lower PTC voltage is required at higher motor voltages due to the internal heat generation of the motor.

To demonstrate integration of separated RPA components, we conduct paper-based isolated RPA experiments with mastermix distributed onto the reaction membrane with magnesium and target added separately. To consistently pipet magnesium and target in discrete and locations on the membrane, we place a clear acrylic (4615T94, McMaster-Carr, USA) laser cut template with two .9 mm circular cutouts spaced 4.5 mm apart (center to center). For these experiments, we pipet 45.5 μL of RPA reaction (mastermix) onto the membrane before aligning the template over the membrane and pipetting 2.5 μL of magnesium acetate and 2 μL of target onto the membrane. Lastly, we remove the laser cut template and place PCR film over the glass slide to seal the membrane.

V. Data collection and analysis

We use an epifluorescence fluorescence microscope (AZ100, Nikon, JPN) and illumination system (X-Cite exacte, Excelitas Technologies, USA) to image the RPA membrane amplification with a 1x objective and field of view diameter of 35 mm. The epifluorescence filter cube set (XF100-2, Omega Optical, LLC., USA) and CMOS camera (Prime BSI Express, Teledyne, Photometrics, USA) captures grayscale images every five seconds up to 20 minutes.

We use an image analysis algorithm (MATLAB, MathWorks, USA) to calculate the fluorescence intensity increase over the full 20-minute image-stack. Background subtraction eliminates artifacts caused by auto-fluorescence due to the membrane, PCR film, glass slide, and initial nonspecific interactions in the assay. The intensity over the entire membrane is averaged and plotted against time. The positive threshold of an RPA reaction is the area averaged intensity of no template control (NTC) HIV DNA experiments plus three standard deviations. All reactions are analyzed for limit of detection at the 15-minute mark.

To examine bulk stirring, we use an analysis algorithm that calculates the standard deviation of averaged fluorescence intensity as a function of time, known as the macro-mixing index. This is given as $\sigma ={\langle {(I-\langle I\rangle )}^2\rangle }^{{}^1∕_2}$, where intensity $I$ is scaled from 0 to 1 with $\langle ⬚\rangle$ indicating the area average intensity in the region of interest. Our region of interest for all experiments is 7.2 mm by 7.2 mm. In this scaling, 0 represents a fully mixed membrane experiment and 1 represents an unmixed experiment. This is the same metric used by Stroock et al. to represent channel mixing for chaotic microchannel mixers.⁵¹ The macro-mixing index is initially set to 1 by dividing the standard deviation over time, σ, when the motor is running by the standard deviation of the first frame, before the mixing motor is turned on, code available in SI.

We determine the z-axis, perpendicular to the motor face, amplitudes produced by the motor using a single-point laser doppler vibrometer (Polytec, DE) to measure the outer edge of the motor face, for each isolation material assessed. Velocity data is recorded using a vibrometer controller (OFV 2600, Polytec, DE) and integrated with respect to time to calculate amplitude data.

To determine the temperature of the RPA reactions, we attach a thermocouple (80BK-A, Fluke Networks, USA) to the motor face and record the data using a digital multimeter (Fluke Networks, USA) every minute. The temperature is monitored without the glass slide, glass fiber membrane, and PCR film due to limited spacing for the thermocouple. When the RPA reaction plate is added to the system, the warmup time to reach incubation temperature will be impacted due to its larger thermal mass; however, we expect this difference to be small as the thermal mass of the glass slide and membrane is relatively low compared to the amount of heat generated from the PTC heater and motor platform. For the motor only case, we remove the PTC heater and place styrofoam under the vibration motor as well as on top of the platform with a central opening to visualize the membrane reaction. Insulation material such as styrofoam and rubber isolation material help retain heat produced from the motor and provide more consistent temperature profiles for the 20-minute duration.

Results and Discussion

We demonstrate that continuous membrane mixing of RPA reactions redistributes reactants and increases amplification regions, thus increasing reaction efficiency and total fluorescence products. To determine the effectiveness of the vibration platform for RPA, we observed and quantified the overall bulk stirring in the system and evaluated various materials and input voltages for the motor to find the most favorable conditions for RPA amplification. We then performed RPA and RT-RPA experiments on glass fiber membranes for 10 – 1,000 cps rxn⁻¹ and 50-1,000 cps rxn⁻¹ respectively, to measure the impact on RPA detection limit, time to threshold, and total fluorescence. We additionally compared 30,000 cps rxn⁻¹ DNA unmixed experiments to 200 cps rxn⁻¹ DNA mixed experiments to observe unmix and mixed scenarios with similar fluorescence curves. We examine conditions with initially homogenized RPA experiments (with and without PTC heater) as well as isolated RPA experiments.

I. Visualization of Bulk Stirring Efficiency

We first study the bulk stirring efficiency in a glass fiber membrane using fluorescently labeled DNA. We pipetted viscous RPA reaction buffer evenly to the membrane and then pipetted a small volume of labeled DNA in the center of the membrane. Using this experimental setup, we can track the efficiency of bulk stirring in the membrane.

Figure 2A depicts images of fluorescently labeled DNA that was pipetted onto a membrane that is saturated with viscous RPA mastermix. Qualitatively, we find that the labeled DNA in the unmixed case does not noticeably diffuse during the 20-minute timespan. In the mixed scenario, we observe rapid mixing with a clockwise bulk flow over the 20-minute timespan. We hypothesize that the clockwise motion is likely due to the rotation of the ERM motor. Figure 2B presents a plot of macro-mixing index versus time for a mixed, unmixed, and homogenized control case. The homogenized case represents a fully pre-stirred control where the concentration is uniform across the entire membrane. In this experiment the labeled DNA is uniformly distributed in the mastermix, via vortexing, before applying the solution to the membrane. We observe bulk stirring efficiency by analyzing the standard deviation of the fluorescence intensity, labeled as macro-mixing index. At a value of 1, the macro-mixing index represents an unmixed experiment and 0 represents a fully mixed experiment. The homogenized sample has a value of nearly zero for the 20-minute timespan. In the unmixed case, the index begins at a value of 1 and marginally decreases as the fluorescent dye diffuses over time. In the mixed case, the area averaged standard deviation originates at an unmixed at a value of 1 and declines rapidly over the first five minutes to a value of 0.35. After 20 minutes, the index reaches 0.1. These results demonstrate effective stirring using vibration-based mixing. The macro-mixing index is indicative of fluid stirring at a length scale finer than the spatial resolution of the imaging system. These experiments do not demonstrate mixing, which requires both stirring and molecular diffusion. We use RPA reactions to provide a clear indication of mixing. RPA requires reagents, magnesium, and target to be mixed at a molecular scale to provide the amplification reaction and resulting fluorescence product. Many mixing demonstrations use colorimetric or fluorescent dyes to represent mixing.^31,32,36,51 These often provide ambiguous metrics that do not necessarily indicate interaction and molecular diffusion between dyes. When using reaction-based assays, such as RPA, we are able to demonstrate clear and unambiguous mixing measurements in the form of a distinct colorimetric or fluorescent reaction product.^52-55

Figure 2.

(A) Fluorescence images of fluorescently labeled DNA within glass fiber membranes that were either unmixed or continuously mixed with a coin motor over a 20-minute period. (B) Macro-mixing index, indicating the stirring efficiency of the platform, as a function of time for a control, mixed, and unmixed conditions. The unmixed state will have a value of 1, and if the system is fully mixed, it will have a value of 0. The red line is an unmixed experiment where minimal mixing due to diffusion occurs over the span of 20 minutes. The mixed experiment in blue indicates a well-mixed experiment by the end of 20 minutes. A 2V input was applied to the motor.

II. Evaluation of vibration mixing platform parameters

To optimize our mixing platform for enhancing RPA reactions, we examined a range of mixing conditions, including various soft isolation materials. The isolation material can be selected to significantly increase the induced vibration amplitude of the analytical membrane by the ERM motor, causing an increase of the reaction efficiency and overall fluorescence output as well as decreasing the time to threshold.

Figure 3A shows the bulk fluorescence as a function of time for RPA experiments with no mixing, mixing with no isolation material, and mixing with an added rubber foam isolation material. When no mixing is used, the fluorescence intensity at 15 minutes reaches around 1.5 units (a.u.). When we apply 2V to the coin cell motor to generate vibration, mixing increases and results in a fluorescence intensity of 7.2 a.u. at 15 minutes, as shown in blue in Figure 3A. Compared to an unmixed reaction, the mixed experiment with isolation foam increases overall fluorescence output by 14.4 times and reduces time to threshold by 5.7 minutes. This data shows that mixing enhances RPA reactions and using isolation materials result in greater mixing compared to a rigid attachment to the plate.

Figure 3.

(A) RPA (200 copies HIV DNA per reaction) results for no mixing (red), mixing with no isolation material (blue), and mixing with rubber foam isolation material (orange). (B) RPA (200 copies HIV DNA) results for various isolation materials. Materials without a voltage label use a 2 volt motor input. For the best performing material, rubber foam isolation material, RPA (200 copies HIV DNA) experiments were also done for various motor input voltages (also seen in Figure S5A). (C) Vibrometer results for z-amplitude at edge of motor for various isolation materials. The z-amplitude motion is a metric used to analyze optimal isolation material. Grey dashed line at 2 volts represents motor input voltage used for majority of the screened isolation materials in Figure 3B. (D) Temperature profiles for the platform at motor inputs from 1-3V. With a PTC added to the system, the 2V motor input increases the temperature profile by 2 °C from 10-12 minutes and the PTC maintains a steady temperature profile within RPA conditions (grey shading).

Figure 3B depicts the RPA time to threshold versus the area average fluorescence at 15 minutes for varying motor inputs and isolation materials. Ideally, RPA reactions will exceed the threshold rapidly (short times) and yield high fluorescence (upper left-hand corner). We screened five off the shelf isolation materials that are used to reduce vibrations to determine which material provided the best RPA results in terms of time to threshold and overall fluorescence product. Compared to the unmixed and no material mixed reaction, the isolation materials significantly improved overall fluorescence and moderately improved time to threshold results. We observed that the silicone gel and PVC based rubber foam materials performed similarly in terms of time to threshold (around 7 minutes) and endpoint fluorescence at 15 minutes (around 13 a.u.). We observed a moderate improvement in time to threshold and endpoint fluorescence for the polyfoam and neoprene material as well. Depending on the assay used, how the platform is configured or integrated into a larger device, or what materials are used for the device, these results will likely change. For our assay, we chose the rubber foam material as it provided the most consistent results while reducing the time to threshold and significantly improving the overall fluorescence.

Figure 3C shows the maximum z-amplitude of the coin motor as a function of the motor driving voltage for five materials. These results indicate that the amplitude increases with increasing driving voltage and that foam materials produce higher z-amplitudes for a given driving voltage. We hypothesize that a low stiffness insulation material acts as a loose spring in a mass-spring system and produces a pronounced rocking motion in the z-axis, perpendicular to the face of the motor, when the ERM motor spins in-plane. As a result of the rocking motion, we expect the outer edge of the motor to have the largest amplitude delta. Out of the materials tested, we found that the rubber foam had a high z-amplitude and provided the best RPA performance in terms of time-to-threshold and fluorescence signal output.

RPA reactions require a temperature of 37-42 °C for optimal reaction kinetics.^28,56 We achieve this in our system using a PTC heater in combination with heat released from the coin cell motor. The PTC element is a self-regulating heater that can maintain a constant temperature. We observed that with the PTC added into the system, we can achieve consistent RPA results. Figure 3D shows the temperature of the motor surface as a function of time. The data demonstrates that when 4 V is applied to the PTC heater and 2 V is applied to the coin cell motor, 38 °C is achieved with little temperature variation when the system reaches its steady state. To ensure all experiments are within the RPA temperature window, we monitor the temperature using a thermocouple before and after each experiment.

We additionally ran mixing conditions with the rubber isolation foam for a motor input range of 1-3 volts, see Figure 3B. For each motor input, we adjusted the PTC input voltage to reach a temperature of 38 °C at the start of each experiment. With higher motor inputs, we expected to achieve more mixing and as a result, an improved fluorescent output. However, we found that with 3 volts applied to the motor with the PTC, it had a much lower endpoint fluorescence than the 2V motor input. The 2 volt motor input with the rubber foam insulation material again had the fastest time to threshold and highest fluorescence intensity. We hypothesize the lower RPA reaction for the 3V motor is due to the difficulties in maintaining ideal RPA temperature at higher motor inputs and possible reduced contact to the RPA membrane due to increased vibration. Our testing indicates that the highest amplification results came from a 2 volt motor input, 0.08 Watts, and using rubber foam insulation material, which was used as the condition for all the other experiments shown in this work. For the 20-minute duration of the experiment, the motor usage is 0.027 Watt-hours or equivalent to ~1.1% of the capacity of two AA batteries connected in series.

III. Limit of Detection for homogenized RPA

Here we study the impact of mixing on the limit of detection of RPA reactions in membranes. In this case, the DNA target, RPA reagents, and magnesium are stirred in a tube, homogenized, and then pipetted onto a membrane. We performed membrane experiments for mixed and unmixed reactions for 10 - 200 copies and 10 - 30,000 copies per reaction, respectively. Figure 4A illustrates the initial distribution of reagents within the membrane and fluorescence images of mixed and unmixed reactions after 20 minutes of incubation at 38 °C. Figure 4B depicts the area averaged fluorescence for RPA reactions as a function of time for a range of DNA target concentrations in mixed and unmixed states. The results show that RPA fluorescence begins to increase above the baseline level after 5-8 minutes. The reaction proceeds until 14 or more minutes before the fluorescence output asymptotes. This data shows that mixing results in reactions occur sooner and to a greater extent due to significant improvement in reaction efficiency.

Figure 4.

(A) Schematic of distribution of RPA reagents, target, and magnesium at the start of the experiment. Fluorescence images for unmixed and mixed results after 20 minutes for 200 copies HIV DNA. Full series of images for mixed and unmixed experiments are found in the supplement Figure S2. (B) Bulk fluorescence as a function of time for unmixed (dashed lines) and mixed (solid lines) reactions. Color indicates specific copy number. Black lines represent the no template controls (NTC) for both mixed and unmixed configurations. For unmixed experiments, 30-30,000 copies per reaction can be seen from (dashed green to dashed orange respectively). For mixed reactions, 12-200 copies per reaction is seen from (solid purple line to solid red line respectively). (C) The area averaged fluorescence after 15 minutes for mixed and unmixed reactions for n = 3. Red represents the unmixed reactions while blue bars represent mixed reactions. 10 and 0 copies per reactions did not amplify for both mixed and unmixed experiments and therefore do not have any noticeable bar.

Figure 4C shows the area averaged fluorescence after 15 minutes for mixed and unmixed reactions with varying input copy numbers. This data shows that the fluorescence is greater for higher input copies and when mixing is employed. The mixed reactions exhibit 5 to 16 times greater fluorescence output than the unmixed cases, depending on respective input copy number. The time to threshold for positive mixed reactions is on average 4.7 minutes less than in the unmixed case. The limit of detection for mixed experiments is 12 copies per reaction with all 3 experiments amplifying. In comparison, the limit of detection for unmixed experiments is 77 copies per reaction, determined using probit analysis (see Figure S1 in SI). To emphasize the degree to which mixing improves paper-based RPA, we included unmixed reactions with very high input copy numbers, and we find that mixed reactions at only 200 copies yield comparable output fluorescence and faster times to threshold than unmixed reactions at 30,000 input copies. In summary, mixing improves the fluorescence output, the time to threshold, and the limit of detection of RPA reactions in porous substrates.

Many viruses, such as HIV, have RNA genomes which are difficult to amplify because RNA is fragile and requires an additional reverse transcription step. To observe if the platform improves single-step reverse transcription RPA, we conducted experiments with and without mixing at 50, 200, and 1,000 HIV RNA copies per reaction. Figure 5 shows RT-RPA fluorescence as a function of time. For 1,000 copies of RNA per reaction, the mixing increased the fluorescence output more than 6 times higher than the unmixed experiments. The time to threshold also decreased by roughly 5 minutes. For 200 copies, endpoint fluorescence output for mixed experiments is more than 20 times higher than unmixed experiments. At a low copy reaction of 50 copies of RNA, all mixed experiments amplified whereas for the unmixed experiment, only two out of three experiments amplified. These experiments correlate well with the DNA experiments and demonstrate improved limit of detection, overall fluorescence output, and time to threshold values with continuous mixing.

Figure 5.

HIV RNA results for 1000, 200, and 50 copies per reaction. Solid lines represent mixed reactions and dashed lines represent unmixed reactions. Black lines represent no template controls (NTC).

IV. Mixing results of isolated RPA: spatially separated magnesium and target

In many POC devices, sample preparation requires integration of purified target into an amplification or detection reaction. As a result, target and other additional components, like magnesium cofactor that is typically added last, may not be well mixed with the rest of the reactants in the assay. RPA reactions in membranes are especially challenging because of the high viscosity and porous membranes. The high viscosity RPA mastermix results in lower diffusion rates when compared to reactions of other NAAT chemistries, such as LAMP. Membranes such as glass fiber, further reduce effective diffusion due to various factors, including its pore structure (porosity and high tortuosity).^57-59 To examine this scenario of spatially separated reactants, we performed isolated RPA experiments with target and magnesium separately applied to RPA reagents, as shown in Figure 6A.

Figure 6.

(A) Depiction of distribution of RPA reagents, target, and magnesium at start of experiment. Fluorescence images for unmixed and mixed results at 20 minutes for 200 copies HIV DNA. (B) Bulk fluorescence results for unmixed reactions (dashed lines) and mixed reactions (solid lines). Color indicates specific copy number. Black lines represent the no template controls (NTC) for mixed and unmixed reactions. (C) Area averaged fluorescence (n=3) at 15 minutes for mixed and unmixed reactions.

Images of membrane isolated RPA reactions for the mixed and unmixed conditions for 200 copies of HIV DNA after 20 minutes of heating at 37 °C, are shown in Figure 6A. The mixed reactions resulted in high fluorescence output while the unmixed reactions did not produce any. This indicates the importance of mixing in reactions that have spatially separated reaction components. Figure 6B presents the bulk fluorescence as a function of time for the mixed and unmixed conditions at various input copy numbers of DNA. The data shows that the average fluorescence for unmixed reactions does not exceed the background level except for one weak reaction at 200 copies per reaction. The background fluorescence for these experiments was less than that of assays that were homogenized with target and cofactor before being deposited on the membrane (see Figures 4 and 5). The low initial background is likely the reason the overall bulk fluorescence is sometimes higher for these experiments than in the homogenized reactions. When mixing is used, RPA amplifies in roughly 6 minutes and increases exponentially, similar to typical in-tube RPA experiments.

Figure 6C depicts the area averaged fluorescence after 15 minutes with n=3. The unmixed case shows almost no fluorescence while the mixed case indicates strong fluorescence that increases with number of copies of target DNA. We tested 30,000 copies for the unmixed scenario to see if increasing copy number would result in amplification; in this experiment, all three experiments did not amplify. This underscores that in isolated RPA reactions without mixing, limited diffusive mixing occurs between target, magnesium, and RPA reagents during the 20-minute incubation. This suggests that for similar scenarios where uneven integration of target and cofactor are introduced, continuous mixing can effectively distribute separate components and significantly improve membrane amplification results.

V. Vibration based mixing with no additional heating

We show that the coin motor can generate sufficient heat to bring the homogenized RPA reactions to the optimal temperature range (37-42 °C). Removing the PTC heater may reduce the cost, complexity, and power requirements for the system. Instead of a heater, amplification reactions could use a mixing module that can both mix and heat a reaction. In Figure 3D, we show that the temperature profile can reach 37-42 °C in the absence of the PTC heater with a driving motor voltage of 2.4 - 2.6 volts, 0.1 – 0.12 Watts. When compared to the system with the PTC, this configuration requires more time to heat to RPA temperatures and has higher temperature profile instability as the motor input is increased.

Figure 7 shows area averaged fluorescence as a function of time for 200 and 1,000 copies of HIV-1 DNA in RPA reactions conducted in membranes with mixing and heat generated by a coin motor. In this demonstration, homogenized RPA is deposited onto a GF/DVA membrane. In unmixed reactions where no heat is generated, there is no amplification seen over the 20-minute duration of the test. All six experiments conducted with the coin cell motor running resulted in positive amplification. These experiments with no PTC heater exhibit lower amplification fluorescence compared to a mixed scenario with a PTC (Figure 4) and have higher variability between tests. This is likely due to the significantly smaller heating area and non-regulating heating, typically provided by the PTC. With the PTC added in the system, the larger PTC area provides a lower surrounding temperature of 36 °C with the motor head having a temperature of 38 °C when turned on (see Supplementary Figure S4). In comparison, heating solely provided by the motor provides a small surface area, approximately 4.4 times smaller than the area of the PTC and does not have any self-regulating heating ability. The PTC adds a larger overall area with sufficient heating and prevents significant temperature fluctuations on the motor face due to its self-regulating behavior, providing more consistent results. Compared to unmixed reactions with PTC heating (dashed lines in Figure 4B), mixing with no PTC still provides significantly improved amplification results. These experiments demonstrate the feasibility of performing mixed RPA without any dedicated heat source.

Figure 7.

Bulk fluorescence plots for 1000 and 200 copies HIV DNA for vibration-based mixing with no PTC. Solid lines represent mixed experiments, dashed lines represent unmixed experiments, and black lines represent no template controls (NTC).

Summary

We present a vibration-based platform to improve RPA reactions in porous membranes. When continuously mixed, we observe a significant improvement in limit of detection, time to threshold, and fluorescence signal output for HIV-1 DNA and RNA. For DNA targets, we report a limit of detection of 12 copies per reaction (6.4 times improvement in LOD), a reduction in time to threshold from an average of 9.93 minutes to 5.17 minutes, and an overall fluorescence output up to 16 times brighter when compared to unmixed experiments. We show that when the target and magnesium cofactor are initially spatially separated, RPA reactions do not occur unless active mixing is employed. Our results also demonstrate amplification without any additional heating module. The mixing device significantly increases endpoint fluorescence in all conditions.

The high viscosity of RPA, due to its PEG crowding agent, and the porosity of the membrane both contribute to lower diffusion rates in paper-based RPA. The high viscosity reduces diffusion due to Stokes-Einstein diffusivity, and the glass fiber membranes reduce the effective diffusivity due to its porosity (ω) and tortuosity.^57,59 In the simplest scaling model for porous media, the effective diffusivity scales as, $D_{eff}={\omega }^{3∕2}D$.⁵⁸ With mixing, we are able to address the lowered effective diffusion and improve reaction efficiency of paper-based RPA. While RPA is the focus of this work, we hypothesize that continuous active mixing may enhance other paper-based isothermal nucleic acid amplification assays, such as LAMP or strand displacement amplification (SDA). We observed that diffusive mixing in glass fiber membranes is limited even in the absence of the PEG crowding agent (see Figure S6); therefore, our mixing platform may improve the reaction efficiency of paper-based NAATs that do not feature highly viscous buffer conditions.

This work shows that mixing modules for membrane-based NAATs can be leveraged for low-cost, point-of-care nucleic acid amplification to provide robust tests with high reaction efficiency and fluorescence output. Effective mixing can improve clarity of results, consistency at lower copy numbers, and further reduce cost in detection equipment by increasing reaction output (i.e. fluorescence). None of the experiments required intermediate user steps after depositing the reaction on the membrane as mixing and heating was triggered remotely and remained on for the entire duration of the experiments. This work shows that reagents that are initially spatially separated (e.g. target and cofactors) can be mixed effectively, which may be a helpful tool when integrating on-chip sample preparation steps with nucleic acid amplification and detection. Future work includes comparing with other mixing methods, comparing vibration devices such as tweeters or PZT transducers, observing output with low-cost sensors and filters, exploring mixing for enhanced paper-based LAMP, and integrating sample preparation with the platform.