## Main

For three decades, polymerase chain reaction (PCR) has served as the gold-standard technique for the detection of nucleic acids from human clinical specimens1,2. However, PCR is not commonly performed as a clinical diagnostic at point-of-care (POC) settings, with one limitation being the difficulty and cost of instrumentation for ramping up and down the temperature in a repeated, fast and controlled manner3,4,5,6. In particular, traditional thermocyclers set the temperature of the reaction vessel using conductive and convective heat transfer from heat blocks that exhibit the Peltier effect; due to large thermal mass and low thermal conductivity of the heat blocks7, the thermocycling procedure requires bulky instrumentation and long operation time. These requirements have limited PCR to mainly the laboratory settings, contributing to a scarcity of accurate POC diagnostics and bottlenecks in test results during the COVID-19 pandemic6,8,9,10,11.

PCR is most commonly offered in the form of quantitative PCR (qPCR), which has transformed clinical diagnostics12 in allowing amplified products to be detected earlier during a run, and for infectious units to be quantified via cycle threshold (Ct) values. Compared with end-point PCR, the real-time fluorescence monitoring feature of qPCR poses additional technical hurdles for POC miniaturization. Instead of heat blocks, our approach for thermocycling uses direct heating of the solution via plasmonic nanoparticles, in which infrared (IR) radiation interacts with the electrons of nanoparticles to produce oscillation of the electron cloud and subsequent generation of heat; this photothermal method achieves rapid heating of solutions13 with small optical components.

Although plasmonic heating has been applied for therapeutic14,15 and molecular diagnostic16,17,18 applications, previous work in plasmonic PCR thermocycling has not demonstrated a real-time fluorescence monitoring capability, a central technical feature of qPCR (as opposed to end-point PCR19), which is the backbone of laboratory-based PCR methods for clinical diagnostics, including COVID-19 (refs. 20,21). For example, a previous study on plasmonic thermocycling examined fluorescence at a single wavelength on purified nucleic acids in buffer16, and another study used plasmonic heating for end-point PCR rather than qPCR, and did not perform multiplexing within a single sample22; further comparison with previous work can be found in Supplementary Tables 1 and 2. A technical limitation to achieve real-time monitoring and qPCR has been fluorescence quenching from nanoparticles. Here we seek to achieve reverse-transcriptase quantitative PCR (RT-qPCR) in a single reaction vessel containing PCR chemistry, fluorescent probes and plasmonic nanoparticles, with the capability of multiplexed fluorescence monitoring during plasmonic thermocycling. We seek to demonstrate RT-qPCR to detect RNA from SARS-CoV-2 from both human saliva and nasal specimens, using small optical components, and more quickly from sample to result than current commercial POC devices (that is, under 30 min).

## Main findings

### Plasmonic RT-PCR with real-time fluorescence monitoring

We used an optics-driven setup for achieving both thermocycling at IR wavelengths and multispectral fluorescence measurements at visible wavelengths (Fig. 1a). A combination of nanomaterial selection, instrument setup and deconvolution software enabled this capability. First, in the reaction vessels, we used gold nanorods (AuNRs) with localized surface plasmon resonance in the near-infrared range (~850 nm); this wavelength range allowed the use of fluorescent probes for real-time fluorescence detection without the need to remove AuNRs (another study22 used nanoparticles with the maximum absorption at 535 nm, which overlaps with the emission spectra of many currently used fluorescent probes for PCR). To drive thermocycling, an optical setup consisting of three IR light-emitting diodes (LEDs), operating at 850 nm and concentrically positioned around a thin-walled PCR tube, produced rapid heating throughout the tube compatible with RT-PCR (Fig. 1b and Extended Data Fig. 1a). With previous work on plasmonic heating showing fluorescence quenching by AuNRs22 and consequently an inability to monitor fluorescence in real time (as a separate mechanical step was needed for each fluorescence recording), we used a concentration of AuNRs that was sufficiently high to achieve the photothermal effect but also would not interfere with fluorescence measurements, to rapidly generate heat throughout the solution (20 µl in volume); during each cycle, cooling was achieved with a small 12 V fan (Fig. 1b and Extended Data Fig. 1a). The intensity and time sequence of the LEDs were calibrated using a K-type thermocouple and custom LabVIEW program, which was fed into a second LabVIEW program to automate the tuning of an open-loop control system.

### 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.