Zephyrnet Logo

Multiplexed reverse-transcriptase quantitative polymerase chain reaction using plasmonic nanoparticles for point-of-care COVID-19 diagnosis

Date:

Design of instrument for plasmonic thermocycling

The integrated setup consisted of both plasmonic thermocycling and multispectral fluorometry, both of which can operate on the same reaction vessel with no moving components or steps. The plasmonic thermocycling prototype was developed with the following capabilities: (1) consistent heating via AuNRs, (2) real-time fluorescence detection and (3) compatibility with a simple sample cartridge. Briefly, a hexagonal three-dimensional printed hub was designed to contain all the optical components concentrically surrounding the PCR tube. Three IR LEDs, arranged in a hexagon-like manner, were aimed at the PCR tube. Each LED was positioned underneath a lens and on top of a custom-machined heat sink and fan to prevent overheating. An additional 12 V fan was positioned near the PCR tube and used to cool the sample. An airflow chamber was cut out of the bottom of the three-dimensional printed hub to ensure airflow over the PCR tube. The fluorometer components (Fig. 1c, laser and spectrometer setup) were positioned in the remaining sides of the hub and connected to an Arduino Mega board. The fluorometer was designed to comprise a 488 nm laser diode as the excitation source, with light passing through a collimating lens and bandpass filter before entering the PCR tube. The light emitted by the reaction fluid is passed through a series of filters (excitation blocking, IR blocking and longpass filtering) and a condensing lens and is then collected into a 600-μm-diameter optical fibre, and measured by a spectrometer (Ibsen). An annotated image of the prototype (Extended Data Fig. 1) further demonstrates the hexagonal positioning of the components, and a wiring diagram (Extended Data Fig. 2) details the electronic connections used.

Reaction conditions

Functionalized silica-coated AuNRs were purchased from Nanopartz. AuNRs had a surface plasmon resonance peak of ~850 nm (slight variation between batches) and an aspect ratio of ~4.5 nm. A 10 nm silica coating was used to prevent the adsorption of proteins during the PCR reaction. Unless otherwise noted, the reaction conditions were as follows. PCR reactions consisted of the following: 10 µl 2X PrimeScript III from Takara (Cat. RR600A), AuNRs (Nanopartz, final OD of 2), 500 nM forward and reverse primers (Integrated DNA Technologies), and 125 nM probes (Integrated DNA Technologies). Similar to the Centers for Disease Control and Prevention’s RT-PCR test, we used the N1 and N2 primers and probes to target the SARS-CoV-2 nucleocapsid gene, and the RP primers and probes to target a human ribosomal protein (Rp) gene (Supplementary Table 3). Some reactions used an alternative nucleocapsid gene (designated ‘N-gene’) (Supplementary Table 3). The targets were detected using a combination of FAM, SUN, HEX and ROX fluorophores. For all the reactions on the prototypes, evaporation was prevented with Chill-out liquid wax (Bio-Rad) or mineral oil, unless otherwise noted.

Unless otherwise noted, for the plasmonic instrument, thermocycling conditions were as follows: reverse transcription for 2–5 min at 45–50 °C, followed by initial denaturation for 10–20 s at 95 °C. Next, the reaction cycled 40–45 times between a low temperature (58–60 °C) held for 0–8 s and a high temperature (91–97 °C) held for 0–1 s.

Initial testing of amplification using plasmonic RT-PCR

For experiments with RNA (Fig. 2), spiked RNA (BEI, Cat. NR-52285) in 1X TE buffer (10 mM Tris-HCl and 1 mM EDTA, Integrated DNA Technologies) was used to bring the reaction volume up to 20 µl. The NTCs were tested with the same mix and conditions but without any SARS-CoV-2 RNA template, which was instead replaced with TE buffer only. Positive and negative controls were also run on a QuantStudio 6 Pro system. The prototype temperature was controlled using a K-type thermocouple for closed-loop thermal cycling conditions (Fig. 2a,b). A Python script identified the times and temperatures at which the maximum and minimum temperatures were reached and then the average heating and cooling rates were calculated for each cycle. Fluorescence measurements were made on a BioTek plate reader by aliquoting the 20 µl post-PCR reaction to a 384-well plate, and the emission data were collected for 5-FAM, SUN and ROX dyes (measuring the amplification of SARS-CoV-2 N1, SARS-CoV-2 N2 and human RP, respectively). The LoD threshold was determined by running three NTCs and taking the mean raw fluorescence plus ten times the standard deviation (Fig. 2d, dotted line). All the concentrations were run in triplicate, except 2,960 copies per millilitre (which had six replicates). Raw fluorescence for each concentration was compared with the NTC value via one-way ANOVA followed by Sidak’s multiple comparisons test.

For the experiments in Fig. 3a,b, a final concentration of OD of 18 was used for the nanoparticles. A 10 µl PCR reaction with TaqPath ProAmp Master Mix, CG, from Thermo Fisher Scientific (Cat. A30865) was used with 500 nM forward and reverse primers (Integrated DNA Technologies) and 125 nM probes (Integrated DNA Technologies). A complementary DNA sample was thermocycled on the QuantStudio system for 20 s at 95 °C followed by 40 cycles between 95 °C (1 s) and 60 °C (2 s). After the cycles, there was a 30 s hold at 60 °C. The Ct values were determined by the QuantStudio desktop analysis software.

For the experiments in Fig. 3c–e, spiked inactivated virus (BEI, Cat. NR-52286) in 1:1 mixture of donor saliva (Innovative Research, Cat. IRHUSLS5ML) and 1X TE buffer was used to bring the reaction volume up to 20 µl. The temperature was controlled with a thermocouple for closed-loop cycling parameters. Fluorescence measurements on the prototype were made as follows: raw fluorescence spectra were collected by the spectrometer at the end of the annealing/extension step during each cycle. These spectra were analysed using least squares regression, based on ideal peaks experimentally determined by measuring the fluorescence spectra for amplicons containing the PCR product from a single fluorophore. Next, each component signal was plotted against the cycle number. For Fig. 3e, the fluorescence for each fluorophore was normalized by its maximum value in that run.

For Extended Data Fig. 3, a 10 µl PCR reaction with Reliance One-Step Multiplex RT-qPCR Supermix from Bio-Rad (Cat. 12010176) and nanoparticles with a final concentration of OD of 18 was used. N1 was detected with FAM and RP was detected with HEX. The final primer concentration was 400 nM and the final probe concentration was 100 nM. Plasmonic thermal cycling conditions consisted of 1 min at 45 °C and 20 s at 95 °C, followed by 40 cycles between 54 °C (20 s) and 95 °C (0 s), as controlled by an IR pyrometer. Evaporation was prevented with 75 µl of Chill-out PCR wax. Saliva specimens were obtained from Mirimus Foundation and SUNY Downstate from patients suspected of being infected with SARS-CoV-2 and stored at 4 °C on receipt. The samples were heat inactivated for 5 min at 95 °C before receipt. The samples were diluted at 1:1 in 1X TE buffer and added to the PCR reaction mix to reach 10 µl. Fluorescence was measured after amplification on a BioTek plate reader.

Characterization of heating and fluorescence quenching by AuNRs

For the AuNR heating-rate characterization (Fig. 3f), dilutions were prepared in 1X TE buffer with AuNR concentrations ranging from OD of 0.5 to 8.0. Five samples of each concentration were thermocycled between 60 and 95 °C eight times (no holds) by a closed-loop LabVIEW program. A Python script identified the times and temperatures at which the maximum and minimum temperatures were reached and then the average heating and cooling rates were calculated for each cycle. The results are reported as the average of 40 measurements per concentration.

For the quenching characterization (Fig. 3g), dilutions were prepared in 1X TE buffer with AuNR concentrations ranging from OD of 0 to 32. An oligo with FAM attached was used as a fluorophore. For each concentration of AuNR, the concentration of FAM oligo was kept constant at 1.43 μM. All the values are normalized to the average fluorescence measurements of the samples with OD of 0 (the brightest sample). For each concentration, three samples were measured three times each, and the data are reported as nine replicates. The optical LoD was calculated as the average plus ten times the standard deviation of stock AuNR (OD of 94).

Evaluation of performance of plasmonic PCR

For Fig. 4a–c,e, spiked inactivated virus (BEI, Cat. NR-52286) in 1:1 mixture of donor saliva (Innovative Research, Cat. IRHUSLS5ML) and 1X TE buffer was used to bring the reaction volume up to 20 µl. The NTCs were tested with the same mix and conditions but without any template (inactivated SARS-CoV-2 virus). Buffer NTCs indicate the use of a TE buffer only, and saliva NTCs indicate the use of donor saliva mixed in 1:1 with the TE buffer as described below. Positive and negative controls were also run on a QuantStudio 6 Pro system. Thermocycling conditions were initially calibrated in LabVIEW using closed-loop control, and the output was converted to an open-loop format for testing. The samples were considered positive on the prototype if both N1 and RP were detected (Ct < 46). As commonly done for diagnostic tests, samples for which human RP failed to amplify were considered indeterminate42. Onboard fluorescence measurements on the prototype were made as previously described. For Fig. 4b, N1 fluorescence was smoothed and normalized. The baseline was individually selected for each run to account for fluorometer noise. The fluorescence threshold value was calculated as the mean plus ten times the standard deviation of the baseline. The Ct value was calculated to be the interpolated cycle at which the fluorescence signal crossed the calculated threshold value. Any Ct value less than the number of cycles was interpreted as positive.

For Fig. 4d, spiked inactivated virus (BEI, Cat. NR-52286) in 1:1 mixture of donor saliva (Innovative Research, Cat. IRHUSLS5ML) and 1X TE buffer was used to bring the reaction volume up to 20 µl. The NTCs were tested with the same mix and conditions but without any template (inactivated SARS-CoV-2 virus). Saliva NTCs indicate the use of donor saliva mixed in 1:1 with the TE buffer as described above. Positive and negative controls were also run on a QuantStudio 6 Pro system. Thermocycling conditions were initially calibrated in LabVIEW using closed-loop control, and the output was converted to an open-loop format for testing. Onboard fluorescence measurements on the prototype were made as described above. The baseline was preselected to be cycles 0–20. The fluorescence threshold value was calculated as the mean plus ten times the standard deviation of the baseline. The Ct value was calculated to be the interpolated cycle at which the fluorescence signal crossed the calculated threshold value. Any Ct value less than the number of cycles was interpreted as positive. The samples were considered positive if N1 or N2 was detected (regardless of RP detection). The LoD was determined as the lowest concentration resulting in the positive detection of at least three out of three samples, based on US FDA guidelines43 and established standard of many products obtaining emergency use authorization.

Detection of SARS-CoV-2 RNA from human clinical saliva samples

For Fig. 4f–h, the PCR reactions were similar to those generally described above. Deidentified clinical specimens were obtained under a protocol approved by the Columbia University Medical Center Institutional Review Board (AAAT0100). Saliva specimens were obtained from Mirimus Foundation and SUNY Downstate from patients suspected of being infected with SARS-CoV-2 and stored at 4 °C on receipt. The samples were not heat inactivated. The samples were diluted at 1:1 in 1X TE buffer and added to the PCR reaction mix (as described above) to reach 20 µl.

Although all the clinical testing occurred within two weeks of receiving the samples, in light of possible sample degradation during transportation and storage, we reconfirmed the status of the specimens by running them in duplicates on laboratory-based qPCR using a Thermo Fisher QuantStudio 6 Pro instrument, which acted as our laboratory-based PCR reference (consistent with previous studies using established RT-qPCR methods to measure SARS-CoV-2 RNA from human saliva specimens44,45). Positive and negative reaction controls were included on every QuantStudio plate: templates for positive control reactions contained 4,425 copies per millilitre of inactivated SARS-CoV-2 virus from nCoV 2019-nCoV/USA-WA1/2020 (BEI, Cat. NR-52286) in 7.9 µl of a single human donor saliva from Innovative Research (Cat. IRHUSLS5ML) mixed in 1:1 with 1X TE buffer from Integrated DNA Technologies (Cat. 11–01–02–02), saliva negative reaction controls contained only saliva and TE buffer, and buffer negative reaction controls only contained the buffer. Reactions were run on a QuantStudio 6 Pro real-time PCR system (Applied Biosystems) using a thermal cycling program for 2 min at 50 °C, 1 min at 95 °C, and 40 cycles of 2 s at 60 °C and 1 s at 95 °C. The Ct values were determined using the QuantStudio desktop analysis software. A sample was considered positive if N1 was detected, negative if N1 was not detected and RP was detected, and indeterminate if neither N1 nor RP was detected (Supplementary Table 4). Indeterminate samples were excluded from the analysis. Only samples for which both duplicate wells were correctly amplified (positives) or did not amplify (negatives) in QuantStudio (for laboratory-based PCR) were considered viable.

For the plasmonic instrument, thermocycling conditions were initially calibrated in LabVIEW using closed-loop control, and the output was converted to an open-loop format for testing. Fluorescence signals were collected and deconvolved as described above. The baseline was preselected to be cycles 8–18 but was modified for two samples to account for fluorescence noise. The fluorescence threshold value was calculated as the mean plus ten times the standard deviation of the baseline. The Ct value was calculated to be the interpolated cycle at which the fluorescence signal crossed the calculated threshold value. Any Ct value less than the number of cycles was interpreted as positive. The samples on the prototype were considered positive if N1 was detected (Ct < 46) regardless of RP detection, negative if N1 was undetected and RP was detected, and indeterminate if both targets were undetected. All the clinical samples were tested once if positive or negative, and twice if indeterminate. If an indeterminate sample tested either positive or negative on repeating the run, that result was used (possibly due to variability of patient saliva and/or an opportunity to explore additional lysis reagents in the buffer, the setup initially produced 14 indeterminate results in which human Rp gene was not initially detected; although re-running the specimen produced viable results for nine samples to leave the total number of indeterminate results to five, we are working to refine the sample collection and preprocessing protocol to minimize the number of indeterminates). For all the viable samples, our results were fully concordant with the reference results from Mirimus Foundation for which the TaqPath kit (emergency use authorization by the US FDA) was used.

Detection of SARS-CoV-2 RNA in human clinical nasal samples

For Fig. 4i–n, deidentified clinical specimens were obtained frozen from the PATH Washington COVID-19 Biorepository (PATH, Seattle, USA), hereafter the Biorepository, and stored at 4 °C after thawing. The samples tested in Fig. 4i–n underwent a 1 min heat lysis step at 95 °C before testing. A 7.7 µl aliquot of undiluted clinical sample was added to each 20 µl PCR reaction. In light of possible sample degradation during transportation and storage, we confirmed the status of the specimens by running them in triplicates on laboratory-based qPCR using a Thermo Fisher QuantStudio 6 Pro instrument, which acted as our laboratory-based PCR reference. QuantStudio reactions used a thermal cycling program of 5 min at 50 °C, 10 s at 95 °C, and 40 cycles of 8 s at 60 °C and 1 s at 95 °C. Positive and negative reaction controls were included on every QuantStudio plate: templates for positive control reactions contained <5,000 copies per millilitre of inactivated SARS-CoV-2 virus from nCoV 2019-nCoV/USA-WA1/2020 (BEI, Cat. NR-52286) in 7.7 µl of 1X TE buffer from Integrated DNA Technologies (Cat. 11-01-02-02), and buffer negative reaction controls contained only buffer and Master Mix. The Ct values were determined using the QuantStudio desktop analysis software. Only samples for which all the three triplicate wells were correctly identified for both N1 and N-gene as positive or negative by QuantStudio were considered viable.

For the plasmonic instrument, the temperatures were controlled using closed-loop sensing with a wire thermocouple, and thermal cycling conditions were programmed in LabVIEW. The Ct values were called using a custom Python script, which identified the fractional cycle at which the second derivative of the fluorescence signal was at its maximum, a method that has been previously validated46. Any Ct value less than the number of cycles was interpreted as positive. A sample was considered positive if N1 or N-gene was detected (regardless of RP detection); negative if neither N1 nor N-gene was detected and RP was detected; and indeterminate if N1, N-gene and RP were not detected (Supplementary Table 5). The samples were tested once if positive or negative, and twice if indeterminate. If an indeterminate sample tested either positive or negative on repeating the run, that result was used. If a sample tested indeterminate a second time on repeating the run, it was excluded from the analysis. Only three samples had indeterminate results. Re-running the samples produced viable results for one of the specimens, leaving the final number of indeterminate samples at two. All the viable samples were concordant with disease status reported from the Biorepository.

Detection of SARS-CoV-2 RNA in diluted human clinical nasal samples

For a subset of samples (Fig. 4o and Extended Data Fig. 4b–e), the QuantStudio Ct value was used to estimate the viral load based on a spiked-virus standard curve. These clinical samples were then diluted in TE buffer to the desired estimated concentration and tested on the plasmonic instrument three times per dilution.

Deidentified clinical specimens were obtained frozen from the Biorepository and stored at 4 °C after thawing. The samples tested in Fig. 4o and Extended Data Fig. 4b–e underwent a 1 min heat lysis step at 95 °C before testing. A 7.5 µl aliquot of undiluted, lysed clinical sample was added to a 20 µl PCR reaction. We first tested the specimens in triplicate on laboratory-based qPCR using a Thermo Fisher QuantStudio 6 Pro instrument. The clinical samples were concurrently run on QuantStudio with a full standard curve, which ranged from 1.8 million copies per millilitre to <5,000 copies per millilitre of heat-inactivated SARS-CoV-2 USA/CA_CDC_5574/2020 (BEI, Cat. NR-55245) spiked into 1X TE buffer. The Ct values were determined using the QuantStudio desktop analysis software.

To determine the viral load of the clinical samples, we first used the spiked-virus serial dilution to correlate the QuantStudio Ct value with the concentration of the spiked virus. A log-plot of the QuantStudio Ct values versus concentration of spiked virus was generated in GraphPad Prism and a linear regression equation was fit to the data (Extended Data Fig. 4a). The Ct values from the clinical samples were then translated into estimated viral concentrations based on the regression equation (Fig. 4o and Extended Data Figs. 4 and 5). Concentrations for diluted samples (Fig. 4o and Extended Data Fig. 4b–e) were calculated based on the estimated raw-sample viral concentrations.

From the estimated clinical-sample viral concentrations, we sought to determine whether our device could detect down to the same LoD of virus in clinical samples as we had previously shown with the spiked virus. To do this, we tested the serial dilutions of clinical samples on our plasmonic instrument. The temperatures were controlled using closed-loop sensing with a wire thermocouple, and thermal cycling conditions were programmed in LabVIEW. To quantify the amplification, the fluorescence threshold was calculated as the mean plus ten times the standard deviation of the average of the baseline, which was predetermined to be cycles 11–23. The Ct value was calculated to be the interpolated cycle at which the fluorescence signal crossed the calculated threshold value. Any Ct value less than the number of cycles was interpreted as positive. A sample was considered positive if N1 or N2 was detected (regardless of RP detection); negative if neither N1 nor N2 was detected and RP was detected; and indeterminate if N1, N2 and RP were not detected (Supplementary Table 5). Indeterminate runs were excluded from the analysis.

The clinical-sample serial dilutions were used to create a plasmonic-instrument standard curve (Fig. 4o and Extended Data Fig. 4c–e) with a regression equation fit to the data in GraphPad Prism.

Specificity testing with viruses closely related to SARS-CoV-2

To evaluate whether our assay could distinguish between SARS-CoV-2 and closely related viruses, we tested higher concentrations of MERS coronavirus (BEI, Cat. NR-50171) and human coronavirus NL63 (BEI, Cat. NR-53530) (Fig. 4q).

Heat-inactivated virus (MERS or NL63) was diluted to approximately 1,875,000 copies per millilitre in TE buffer. Then, 7.5 µl of the diluted virus was added to each 20 µl PCR reaction. Each reaction also had spiked total RNA control (Human) (Thermo Fisher, Cat. 4307281) at a concentration of 0.1875 ng µl–1 for RP detection as a sample-processing control. The samples were tested on the plasmonic instrument using closed-loop sensing with a wire thermocouple, and thermal cycling conditions were programmed in LabVIEW. To quantify the amplification, the fluorescence threshold was calculated as the mean plus ten times the standard deviation of the baseline (cycles 11–23). The Ct value was calculated to be the interpolated cycle at which the fluorescence signal crossed the calculated threshold value. Any Ct value less than the number of cycles was interpreted as positive. A sample was considered positive if N1 or N2 was detected (regardless of RP detection); negative if neither N1 nor N2 was detected and RP was detected; and indeterminate if N1, N2 and RP were not detected (Supplementary Table 5). Indeterminate runs were excluded from the analysis.

Sample cartridge for sample-to-result workflow

We designed a sample cartridge that was simple to use and measured out a preset quantity (10 µl) of the specimen into the PCR tube (Fig. 5). This design features a twist cap to aspirate a designated volume of the solution, and a push-to-dispense feature to eject ~10 µl fluid into the reaction mix without allowing for potential sample exposure to the operator (Fig. 5a). First, the user collects approximately 500 µl of saliva into a tube prefilled with 500 µl of 1X TE buffer to achieve 1:1 dilution (~1 min) (dilution of specimen with plain buffer followed by the direct amplification of crude lysate, with no separate step for RNA extraction, has been shown to previously work for nasopharyngeal swabs25; future work on the use of alternate single-step buffers that can lyse host cells in saliva specimens is needed to improve the results). Second, a custom-designed sample cartridge is used to measure 10 µl of the specimen into a PCR tube containing Master Mix (primers, nucleotides, enzymes and AuNRs, as described above) (20 s) (Fig. 5b and Supplementary Video 1). We used a one-step RT-PCR mix (described above) for a single reaction. Third, the user inserts the PCR tube into the reaction module, removes the reaction module holder, inserts the cartridge into the device and starts thermocycling. The cartridge itself sits passively and plays no other role once inserted; it is removed at the end of the test. As in laboratory-based real-time PCR, the measurements are taken after each cycle, and multispectral fluorescence is computed over the duration of 45 cycles. A test result is determined by comparing the Ct values to previously defined thresholds. Including sample preparation steps, our workflow from sample collection to test result consists of three steps from a user and takes place in 22–23 min (Fig. 5c).

For Fig. 5e, Master Mix, AuNR and clinical samples were the same, as described above (Fig. 4f–h). The thermocycling conditions were initially calibrated in LabVIEW using closed-loop control, and the output was converted to an open-loop format for testing. The fluorescence collection and Ct value interpretation were the same as described for Fig. 4f–h, except that the baseline was preselected to be cycles 5–25 for all the samples. The samples that were indeterminate were not retested due to a limited number of cartridges.

Overall, the cost of goods for a test kit, which included the materials and reagents in the sample cartridge and PCR tube, was less than US $10 at scale.

Statistics

All the statistics, including one-way ANOVA followed by Sidak’s multiple comparison tests and one-way ANOVA followed by Tukey’s multiple comparison tests, were performed using GraphPad Prism 9 software.

Informed consent

The PATH Washington COVID-19 Biorepository (PATH, Seattle, USA) is a specimen biorepository that was constructed by adhering to a governance plan with oversight and approval from PATH Legal Services and the PATH Office of Regulatory Affairs to ensure ethical compliance. The nasal eluate samples obtained from the Biorepository were deidentified clinical discard specimens acquired from CLIA registered laboratories that were testing for SARS-CoV-2 using US FDA EUA RT-PCR assays. The clinical saliva samples were deidentified clinical discard specimens obtained from the Mirimus Foundation with patient consent. Both sources of samples were tested in adherence to US FDA guidelines.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

spot_img

Latest Intelligence

spot_img