Non‐invasive adhesive patch microRNA assay recapitulates tissue biomarkers for melanoma

Dear Editor: We show that microRNAs (miRNAs) are substantially more stable than long RNAs in tape-stripped skin samples and can be robustly quantified by digital polymerase chain reaction (dPCR). Using this approach, a miRNA ratio previously validated in tissue specimens retains its ability to distinguish melanocytic nevi (MN) from invasive melanoma (IM) when measured non-invasively. Many clinically suspicious pigmented skin lesions are biopsied to exclude melanoma, yet the vast majority prove to be benign (e.g. MN), resulting in substantial over-biopsy and patient burden1. This reflects an unmet need for objective, molecular pre-biopsy triage tools to help prioritise lesions for histopathologic evaluation. Existing non-invasive approaches, including image-based and RNA-based tape stripping (TS) assays, have shown promise but are limited by variability, technical failure rates, low specificity, or dependence on degradation-prone RNA species.2, 3 A barrier to broader translation has been the lack of molecular biomarkers that are both biologically informative and analytically robust when collected non-invasively. TS uses an adhesive patch to collect stratum corneum material for analysis (Figure 1A). Current TS RNA assays are designed for use by dermatologists to help prioritise ambiguous pigmented lesions2 (Figure 1B). These assays suffer relatively high technical failure rates—up to ∼14% of tests yield invalid results3—likely due to degradation of the long RNA targets. As miRNAs are inherently more resistant to degradation than mRNAs4, we hypothesised that using more stable short RNA biomarkers could improve the reliability of TS-based tests. We first determined that intact miRNAs are collected in TS samples and can be sensitively quantified by dPCR. TapeStation analysis confirmed that long RNA was largely degraded in TS-derived RNA (Figure 1C). Small RNA sequencing of the preserved fraction predominantly mapped to annotated miRNAs (Figure 1D,E). Several miRNA expression ratios (miR-21-5p:miR-125b-5p and miR-21-5p:miR-211-5p) were previously discovered in tissue specimens and later confirmed across large meta-analyses to consistently distinguish melanoma from nevi.5-7 Those studies were performed on miRNAs isolated from formalin-fixed, paraffin-embedded tissue sections from archival specimens. Here, we evaluate whether these previously established miRNA ratios remain diagnostic when measured non-invasively using TS samples. We developed a multiplex dPCR assay to quantify TS-derived miRNAs, using TaqMan Advanced miRNA probe/primer sets for three targets (miR-125b-5p, miR-21-5p, and miR-211-5p). An input dilution series for the miR-125b-5p dPCR assay revealed a limit of quantification (LOQ) of approximately 9.62 pg RNA, corresponding to ∼7.3 copies/µL (cp/µL) (Figure 1F). Outputs for miR-125b-5p and the commonly used loading control miR-16-5p were linearly related over a wide range, supporting the use of miR-125b-5p as a reference (Figure 1g). We excluded the miR-211-5p assay from quantitative analysis because its endpoint dPCR amplitude distribution lacked a discrete positive cloud, precluding objective thresholding and Poisson-based quantification, consistent with inefficient assay performance rather than biological absence of signal (Figure 2). We prospectively collected TS samples from 177 clinically suspicious pigmented lesions (in 168 patients) immediately before biopsy (Table S1). Each TS sample's miRNA profile was later matched to the lesion's histopathologic diagnosis, which included 80 MNs, 36 IMs, and 61 melanomas in situ (MIS). Patient age, sex, and anatomic distribution were consistent with expected clinical patterns across diagnostic categories (Figure 3A,B). Signals exceeded the LOQ in 96.61% of specimens (171/177) and did not correlate with cohort features (Figure S1). Consistent with prior tissue-based studies, the miR-21-5p:miR-125b-5p ratio was significantly higher in IM than in MN (p < 0.0001; Figure 3C). Receiver operating characteristic analysis demonstrated discrimination between IM and MN with an area under the curve (AUC) of 79.6% (95% bootstrap confidence interval [CI] 69.3%–89.8%) (Figure 3D). Using repeated 10-fold cross-validation (200 iterations), a cut point yielded a sensitivity of 69.4% (95% CI 63.9%–75.0%) and specificity of 80.0% (95% CI 80.0%–82.5%) when sensitivity and specificity were equally weighted. Weighting sensitivity twice as important increased sensitivity to 75.0% with reduced specificity (66.3%). MIS, which had not been previously evaluated using this ratio, showed intermediate values between MN and IM, differing significantly from MN, while remaining lower than IM (Figure 3C). When IM and MIS were combined, discrimination from MN was preserved (Figure 3E, AUC = 70.0%, 95% CI 62.3%–77.9%), with expected trade-offs between sensitivity and specificity depending on weighting. In the subset of samples with sufficient RNA yield to perform both assays, miRNA ratio performance compared favourably to TS mRNA markers (PRAME and LINC00518), demonstrating higher sensitivity and specificity and reinforcing the analytic robustness of the miRNA-based approach (Figure 3F,G). To assess assay robustness under extreme handling conditions, matched triads of tape-strip specimens from adjacent skin were either immediately processed or processed after 3 days at room temperature with and without 12 hours at high temperature (HT, 49°C) (Figure 3H). Resulting miR-21-5p:miR-125b-5p ratios showed acceptable precision (mean CV = 11.99%), indicating stable miRNA ratio signal preservation across these conditions. Conclusion: We demonstrate that a TS patch can collect stable miRs from skin lesions and that a miR-21-5p:miR-125b-5p ratio derived from these samples differentiates melanoma from MN with good discriminatory performance. This non-invasive approach recapitulates a melanoma-associated biomarker previously established in tissue within a real-world prospective cohort, supporting its potential utility for prioritising lesions for biopsy rather than for definitive diagnosis. Performance estimates were derived using conservative cross-validation, and confidence intervals remain relatively wide, reflecting the sample size and underscoring the need for larger, independent cohorts to further refine these metrics. The present study focuses on melanocytic lesions, which represent the primary source of diagnostic ambiguity prompting melanoma biopsy, rather than non-melanocytic skin tumours that are typically clinically distinct. While digital PCR provides high analytic sensitivity and precision, future validation studies will need to consider factors such as cost, throughput, reproducibility and platform selection. Importantly, the reported sensitivity indicates that not all melanomas would be identified by this approach, emphasising that it is not intended to exclude melanoma or replace biopsy, but rather to inform prioritisation of lesions for timely clinical evaluation. With further validation, such a strategy could help improve early melanoma detection, particularly in settings with limited access to dermatologic care. Conceptualization: DEKKER C. DEACON, DOUGLAS GROSSMAN and ROBERT L. JUDSON-TORRES. Formal analysis: CARLY A. BECKER, ROBERT L. JUDSON-TORRES, KENNETH M. BOUCHER and CHRIS J. STUBBEN. Funding acquisition: DOUGLAS GROSSMAN and ROBERT L. JUDSON-TORRES. Investigation: CARLY A. BECKER, DAVIDW.N. MWELWA, AMANDA Z. JIANG and ANNABELLE HUNTSMAN. Methodology: DOUGLAS GROSSMAN and ROBERT L. JUDSON-TORRES. Supervision: ROBERT L. JUDSON-TORRES. Original draft preparation: ROBERT L. JUDSON-TORRES. Writing—reviewing and editing: CARLY A. BECKER, ERIC A. SMITH, DEKKER C. DEACON, DOUGLAS GROSSMAN and ROBERT L. JUDSON-TORRES. We would like to thank the patients who consented and participated in this study. We thank Tristan Smart and Lisa West-Kincaid for assistance in collecting TS samples. We acknowledge the DNA Sequencing Core at the University of Utah and thank Derek Warner for their assistance. We acknowledge the High-Throughput Genomics Core at Huntsman Cancer Institute and thank Brian Dalley for their assistance. This work was supported by the National Cancer Institute (R01CA229896 to Robert L. Judson-Torres), the V Foundation (Translational Research Award T2022-007 to Douglas Grossman), and by direct support from the Huntsman Cancer Institute Melanoma Research Centre. Additional support was provided by the National Cancer Institute of the National Institutes of Health under Award Number P30CA042014 and the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number T32TR004392. The authors declare no conflict of interest. This study was approved by the University of Utah IRB (protocol #156786). All subjects provided written informed consent prior to participation. All primary data in the study are available upon request from the corresponding author. The sequencing data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE286566. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.