Photodynamic therapy reduces the burden of small ultraviolet-induced epidermal clones in human and mouse skin

Abstract

Actinic keratoses (AKs) and keratinocyte carcinomas (KCs) arise from prolonged UV exposure, with precursor UV-induced clonal mutations (CMs) appearing in sun-damaged skin. Photodynamic therapy (PDT) is a common field treatment for AKs and early KCs, but its impact on subclinical CMs is unknown. This study examines CMs using targeted ultra-deep sequencing on epidermal samples. By comparing skin before and after PDT in five patients and a mouse model of chronic UV carcinogenesis, a significant reduction in low-frequency mutations post-treatment was revealed. These findings highlight PDT’s potential in modifying subclinical damage and propose low-variant allele frequency CMs as biomarkers for field treatment efficacy.

Dear Editor, Actinic keratoses (AKs) and keratinocyte carcinomas (KCs) arise after decades of ultraviolet (UV) exposure in sun-exposed skin. The precursors of AKs and KCs are keratinocyte cell groups harbouring clonal mutations (CMs) in damaged skin.^1,2 Field treatments can eliminate AKs and early KCs, and reduce cancer risk.³ Photodynamic therapy (PDT) is the only widely available modality to treat patients in the clinic. PDT effectively treats AKs and early KCs, and reduces the risk of KCs progressing.^4,5 Although a clinical benefit of PDT is suspected in skin without visible lesions, its effect on CMs has yet to be examined objectively.

We, and others, have developed sequencing-based methods to detect CMs in healthy epidermis.^1,2,6,7 We also found that CMs with the mutant allele present in < 1% of the measured DNA sequences for a specific mutation [i.e. low variant allele frequency (VAF)] is correlated with KC burden.² In the current study, CMs before and after PDT were characterized and compared to understand the effect of PDT on subclinical field damage. Forty clinically healthy skin samples were analysed; the samples were taken from five patients undergoing red light 5-aminolaevulinic acid (ALA) PDT to treat AKs. Three patients received PDT on the face, one on the forearm and one on the hand. Four 7-mm² epidermal samples were acquired from each patient from PDT-treated areas before and 31–35 days after PDT (Figure 1a). The 40 epidermal DNA samples were sequenced with ultra-deep targeted sequencing (UTS) using a customized panel (237 kb) to capture the most frequently UV-mutated genomic regions. Median coverage was 32 336 times (range 23 623–41 180 times). Mutation profiles were compared before and after PDT. One sample failed sequencing (‘PT2-Sp2’, average sequencing depth = 2746 times) and was excluded. In the remaining 39 samples, 6853 CMs were identified [median number of CMs per sample: 166 (range 45–418)]. Most CMs (n = 5439; 79.4%) matched established patterns of UV signature mutations (USMs).⁸ The number of CMs was significantly reduced between the pre- and post-treatment samples [median 188 (range 52–418) and 151 (range 45–277) per sample, respectively; P = 0.02]. This difference was mostly driven by the low VAF CMs [median 112 (range 38–330) and 78 (range 29–160), respectively; P = 0.03] before PDT vs. after PDT (Figure 1b) but not by high VAF CMs [VAF > 1%; median 76 (range 14–152) and 67 (range 15–154), respectively (P = 0.20)]. Two of the three patients with facial AKs experienced the largest reduction in low VAF CMs (patient 1 = 56%; patient 4 = 40%), whereas the patient with AK on the hand experienced the lowest reduction (6%).

Figure 1.

Photodynamic therapy (PDT) reduces low variant allele frequency (VAF) clonal mutations in human and mouse skin. (a) Human red light 5’aminolaevulinic acid (ALA) PDT skin samples. Eligible adult patients (n = 5) with severely sun-damaged skin and multiple actinic keratoses (AKs) on the face (n = 3), forearm (n = 1) or hand (n = 1) had 6-mm diameter punch biopsies before and 31–35 days after PDT from clinically healthy-appearing skin. The epidermis was separated with enzymatic digestion by Dispase II and cut into four equal pieces. DNA was isolated from these four 7-mm² epidermal sheets before and after PDT. For PDT, the treatment sites were debrided by light curettage and 20% ALA was applied under occlusion for 3 h. The area was irradiated with water-filtered infrared A light (Hydrosun^® 501 halogen lamp with a 4-mm water cuvette at 250 mW cm^–2 total irradiance intensity, water-filtered spectrum 590–1400 nm) for 20 min. Image created with BioRender. (b) Reduced low VAF mutation burden in five patients treated with red light ALA PDT. Red indicates before treatment (‘Pre’) and blue after treatment (‘Post’). Horizontal black bars depict the median low VAF mutations of all samples collected from one patient at one timepoint. (c) Epidermal samples from ultraviolet B (UVB)-exposed and PDT-treated mice. SKH1 hairless mice were exposed to UVB (120 mJ cm^–2) three times weekly for 8 weeks, followed by red light ALA PDT to treat early field disease without visible signs of carcinogenesis. We collected 40 punch biopsies (2–3 mm) from the PDT-treated and untreated regions of 5 mice and 35 punch biopsies (2–3 mm) from 5 UV irradiated, unmatched, non-PDT-treated controls. The epidermis was separated with Dispase II and the 115 isolated DNA samples underwent ultra-deep targeted sequencing of UV mutation-prone regions in the mouse genome. Image created in BioRender. (d) Reduced low VAF mutation burden in SKH1 hairless mice treated with ALA PDT. Each dot depicts the number of low-VAF mutations identified from one sample. All samples are grouped into three conditions: PDT collected from the treated side; matched control from the contralateral untreated side of the back of the mice; and unmatched controls collected from the untreated side of unrelated animals. F, female; M, male; PT, patient.

These findings were validated in SKH1 hairless mice with chronic (three times weekly) UVB exposure (120 mJ cm^–2). Mice were subjected to red light ALA PDT after 8 weeks of UVB treatment when there was early field disease but no visible sign of tumorigenesis. The following samples were collected: 40 epidermal punch biopsy (2–3 mm) samples from 5 animals that underwent PDT, 40 matched control samples from the untreated sides of the same animals and an additional 35 unmatched control samples from 5 unrelated animals (Figure 1c). The 115 epidermal samples were processed for UTS analysis using a customized panel targeting common UV-mutated regions in the mouse genome (size 251 kb) to a similar sequencing depth (median depth = 31 719 times). Following our previously detailed quality-assurance process,² 103 samples met the criteria: 38 PDT-treated, 37 matched control and 28 unmatched control samples. In total, 23 693 mutations were identified, 87.2% of which were USMs. Notably, mouse epidermis samples predominantly exhibited low-frequency mutations (96%) at a much higher rate than in human skin (59%), indicating differences in clone size relative to the amount of tissue biopsied. This difference was likely due to the mode of UV exposure and characteristics of mouse skin. Notably, PDT-treated skin samples exhibited significantly fewer low VAF mutations (median 168 mutations per sample) compared with the matched control samples (median 223; P = 0.009) and the unmatched control samples [median 249; P = 0.002 (Figure 1d)].

PDT is a widely used procedure often directed at large fields with areas of clinically normal skin, but the effect of PDT in these areas has not yet been studied. Our data provide the first direct evidence in clinical samples and a mouse model of the effectiveness of PDT in modifying the subclinical CM burden, and suggests that low VAF CMs are a potential biomarker for treatment efficacy. Although the sample size of the current study was limited, we have demonstrated changes in keratinocyte mutational profiles post-PDT for the first time, and provide objective evidence for the utility of CM assessment in evaluating field treatment response.

Funding sources

The current work is mainly supported by the United States (US) National Institutes of Health (NIH) grant numbers R01CA255242 (awarded to L.W. and G.P.) and partially by NIH MSCP SPORE (1P50CA254865-01A1 to H.M. Zarour and J.M. Kirkwood) Developmental Research Program funds (to G.P., E. Repasky and L.W.). The Bioinformatics and Genomics Shared Resource, Experimental Tumor Model Shared Resource and Comparative Oncology Shared Resource at Roswell Park Comprehensive Cancer Center were supported by the US National Cancer Institute (NCI) grant P30CA016056 (to C.S. Johnson). L.W. was supported, in part, by U24CA274159 (to A. Hutson, S. Liu, M. Morgan and D. Goodrich). S.F., B.O. and D.G. were supported by R25CA181003 (to P. Hershberger). The content is the sole responsibility of the authors and does not necessarily reflect the official views of the funding agencies. The funders had no involvement in the study design, data collection and analysis, decision to publish or preparation of the manuscript.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Ethics statement

The Hungarian National Ethical Committee approved the trial (certificate number: 17062-5/2019 EÜIG), and the Roswell Park Institutional Review Board approved the sequencing and the data analysis.

Patient consent

Patients provided prior written informed consent to participate in the trial.