Quantitative isothermal amplification on paper membranes using amplification nucleation site analysis

Abstract

Quantitative nucleic acid amplification tests (qNAATs) are critical in treating infectious diseases, such as in HIV viral load monitoring or SARS-CoV-2 testing, in which viral load indicates viral suppression or infectivity. Quantitative PCR is the gold standard tool for qNAATs; however, there is a need to develop point-of-care (POC) qNAATs to manage infectious diseases in outpatient clinics, low- and middle-income countries, and the home. Isothermal amplification methods are an emerging tool for POC NAATs as an alternative to traditional PCR-based workflows. Previous works have focused on relating isothermal amplification bulk fluorescence signals to input copies of target nucleic acids for sample quantification with limited success. In this work, we show that recombinase polymerase amplification (RPA) reactions on paper membranes exhibit discrete fluorescent amplification nucleation sites. We demonstrate that the number of nucleation sites can be used to quantify HIV-1 DNA and viral RNA in less than 20 minutes. An image-analysis algorithm quantifies nucleation sites and determines the input nucleic acid copies in the range of 67-3000 copies per reaction. We demonstrate a mobile phone-based system for image capture and onboard processing, illustrating that this method may be used at the point-of-care for qNAATs with minimal instrumentation.

figure 1:. process flow for amplification nucleation site quantification and analysis.

(a) recombinase polymerase amplification mastermix and target are dispensed onto paper membrane. the membrane is covered with pcr tape and heated to 39 °c for 20 minutes. (b) discrete amplification nucleation sites begin to form and are recorded via fluorescent microscopy. (c) the images are analyzed by an image analysis algorithm which quantifies the number of amplification nucleation sites. this value is related to the original sample target concentration.

figure 2:

mobile phone experimental setup for nucleation site analysis, showing (a) mobile phone, petri dish containing the amplification membrane, heater, and 3d-printed fixturing, which holds the (b) excitation and emission filters, as well as the plano-convex macro lens. (c) onboard processing of the images is performed using a custom app which displays the number of sites counted and calculates the estimated nucleic acid copies present.

figure 3:

representative fluorescence images of rpa amplification of hiv-1 dna on gf/dva membrane captured using a microscope and mobile phone with copy number ranging from 30 to 100,000 cps/rxn. images were recorded at t = 750 s. at lower copy numbers (30–3,000 cps/rxn), we observe discrete amplification nucleation sites, with a positive relationship between number of nucleation sites and input copy number. at high copy numbers (10,000 and 100,000 cps/rxn), the nucleation sites are numerous and they merge into a splotchy heterogenous fluorescence image, making quantification difficult.

figure 4:. results of algorithmic nucleation site identification and counting.

(a) circular hough transform (cht) algorithm process performed on microscope-acquired images. the raw data (1) is first resampled to increase pixel density (2) with the pixel intensity averaged over 10 frames (3). we perform background subtraction (4), subtracting the average of the first two minutes’ frames on a pixel-by-pixel basis with the resulting amplification nucleation sites then identified and quantified via a circle counting algorithm (5). (b) comparison of the cht (○) and tap (□) algorithms against a manual count of rpa amplification nucleation sites with dna target copy numbers of 30 – 10,000 cps/rxn. at lower copy numbers (≤ 3,000 cps/rxn), there is good agreement between the manual count and both algorithms. in this regime, the number of amplification nucleation sites is relatively low and there is sufficient separation between the sites for successful algorithmic quantification. at higher copy numbers (10,000 cps/rxn), the nucleation sites begin to merge, making algorithmic quantification more difficult and resulting in an undercount relative to the manual count. this undercount is observed for both algorithms, though it is more pronounced in the tap algorithm. (c) representative experiment of 1,000 cps/rxn dna on whatman gf/dva membrane, showing the number of amplification nucleation sites as a function of experiment time, quantified via the cht image analysis algorithm. a sharp increase in number of nucleation sites occurs near 4–10 minutes, followed by a decrease in number of identified sites as the sites begin to grow and merge. the dashed line represents the moving-mean smoothing function used.

figure 5:

quantification of amplification nucleation sites for hiv dna (blue open circles) and rna (red filled circles) on gf/dva membrane captured using a microscope compared to hiv dna data captured and processed using a mobile phone (yellow filled squares). data points represent the average and standard deviations (n=3) for each target copy number tested and quantified using cht algorithms. dashed lines represent fitted models in all cases, we observe a strong relationship between nucleic acid input copy number and number of amplification nucleation sites. at high copy numbers (>10,000 cps/rxn), the number of measured nucleation sites decreases due to merging of amplification sites and inability of the experimental methodology to resolve and quantify the individual nucleation sites.