Quantitative Detection of Severe COVID-19 in Children via Multiplex Salivary miRNA Profiling Using Ligation-RPA

Introduction: : Pediatric SARS-CoV-2 infections often go unnoticed because children with mild or silent disease are seldom tested, yet a significant minority progress to multisystem inflammatory syndrome in children (MIS-C), acute respiratory distress syndrome, thrombo-inflammation, or cardiac injury. Pinpointing high-risk cases early is therefore essential for timely and effective care. Saliva offers an ideal diagnostic matrix: it can be self-collected without needles and is rich in microRNAs (miRNAs) that reflect systemic inflammation. However, conventional miRNA assays, including northern blotting, microarrays, and RT-PCR, though precise, are slow, costly, and equipment-intensive, limiting their use outside specialized laboratories. Recombinase polymerase amplification (RPA) alleviates many of these barriers by operating at 37-42 °C, generating results rapidly, and accommodating multiplex workflows with minimal primer sets. Yet direct miRNA analysis is hampered because RPA prefers amplicons longer than 80 bp. We addressed this limitation by incorporating a ligase step that first bridges the short targets, enabling femtomolar sensitivity while preserving simplicity. Here we present a portable ligation-RPA analyzer designed to quantify salivary miR-1273, miR-296, and miR-29, biomarkers linked to severe pediatric COVID-19. The system provides simultaneous, quantitative readouts that align clinically meaningful femtomolar concentration ranges. We validated the specificity of the system with both individual and mixed samples, demonstrating that our portable extraction method exhibited performance comparable to benchtop systems. Importantly, the platform differentiated severely from mild infections directly in raw saliva by detecting the characteristic down-regulation of these miRNAs in high-risk children. Our platform offers significant potential for early disease diagnosis and practical application in Point-of-care (POC).

Materials and

Methods: : Portable analyzer & centrifuge. The portable reader was modeled in PTC Creo, 3D-printed on a MakerBot Method X, and assembled with OSH Park PCBs hosting an MCU, MOSFET-regulated heater, in-block thermistor, and an optics board with a trimmer-set blue LED and AS7341 sensor. Control and data transfer run through a Bluetooth Android app. A matching 3D-printed centrifuge houses a 12 V DC motor, Arduino Nano, and 28.8 Wh Li-ion pack.
Sampling & extraction. Saliva from emergency-department patients ( < 18 y, Mar 2021–Feb 2022) was swab-collected, preserved with Genotek reagent, stored at room temperature, and shipped monthly. miRNA was isolated on-site using the miRNeasy Mini Kit at 6000 rpm in the portable centrifuge.
Molecular workflow. Primers and probes for miR-1273, -296, and -29 were designed with PrimerQuest. Hybridization (8 µL, 100 nM targets) was heated to 85 °C for 3 min then iced. Ligation (2 µL) with SpintR ligase and ET SSB ran 15 min at 37 °C and was mixed 1:4 with annealed products. RPA (25 µL) contained 5 µL ligation products, primers, buffer, E-mix, dNTPs, core enzymes, SYTO-9, and MgOAc; amplification ran at 39 °C for 20 min with fluorescence measured every 30 s.
Gel. Ligation and RPA products were resolved on 5 % agarose, stained with SYBR Safe, and imaged on a GelDoc Go. Data were processed in Python and Origin; thresholds were μ + 3σ of triplicate NTCs, positives were defined by time-to-threshold (Tt), and concentration differences assessed by t-test (p < 0.05) with linearity R2.

Results, Conclusions, and Discussions:: Assay validation. To verify assay performance, miRNA was first hybridized with probes P1/P2, ligated at 37 °C for 10 min to create a 5′-phosphorylated ssDNA template, then isothermally amplified by RPA with SYTO9 fluorescence reporting (Figure 1a). Agarose analysis confirmed efficient ligation: the high-molecular-weight band eclipsed individual miR-1273 and probe lanes (Figure 1b). Crucially, Figure 1c compares reagent permutations: only reactions containing miR-29, both probes, primers, and full RPA mix yield the expected ~100 bp amplicon, whereas partial mixes produce faint primer-dimer smears, verifying that specific ligation products drive efficient, target-dependent, robust amplification.
Analyzer sensitivity and specificity. To enable POC miRNA testing, we engineered a palm-sized, eight-channel analyzer: a 3-D-printed case houses a Raspberry Pi PCB, MOSFET-heated aluminum block (39 °C) with thermistor control, blue LED/AS7341 optics, 9 V power, and Bluetooth phone interface (Figure 2a,b). The filter-free layout minimizes crosstalk and noise for real-time fluorescence. Sensitivity was assessed with synthetic miR-1273/ 296/29 from 1 fM to 100 pM: Tt fell monotonically with concentration, and linear fits yielded R² values of 0.90, 0.85, and 0.97, confirming quantitative response across five orders of magnitude (Figure 2c). Specificity testing showed that only cognate probes, or the triplex mixture, generated Tt signals; off-targets and NTCs remained silent, showing 100 % multiplex specificity (Figure 2d).
Benchmarking and miRNA profiling. To benchmark our portable system, Figure 3a traces the saliva-to-result path: heat-inactivated saliva is split for miRNA extraction on a benchtop or portable assembly, then quantified. Portable Tt values aligned closely with benchtop data for miR-1273/ 296/29 (r = 0.95–0.97, R² = 0.85–0.94; Figure 3b), confirming analytical parity. Deploying the workflow on 154 pediatric samples revealed pronounced down-regulation of all three miRNAs in severe cases (Figure 3c). Scatter plots show significant differences between non-severe and severe cohorts (***p < 0.001; Figure 4d–f). Receiver-operating curves yielded AUCs of 0.91, 0.88, and 0.87 for single miRNA and 0.98 when combined, underscoring the advantage of multiplex analysis for severity stratification (Figure 4g,h). Collectively, these data validate the sensitivity, and clinical utility for POC assessment of portable platforms.

Acknowledgements and/or References (Optional):: This work was partially supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health through the RADx program (1R61HD105610), the National Institutes of Health (R33HD105610, R33AI147419), and the National Science Foundation (2319913). Any opinions, findings, conclusions, or recommendations expressed in this work are those of the authors and do not necessarily reflect the views of the National Science Foundation and National Institutes of Health.