Abstract

In dermatology, deep learning may be applied for skin lesion classification. However, for a given input image, a neural network only outputs a label, obtained using the class probabilities, which do not model uncertainty. Our group developed a novel method to quantify uncertainty in stochastic neural networks. In this study, we aimed to train such network for skin lesion classification and evaluate its diagnostic performance and uncertainty, and compare the results to the assessments by a group of dermatologists. By passing duplicates of an image through such a stochastic neural network, we obtained distributions per class, rather than a single probability value. We interpreted the overlap between these distributions as the output uncertainty, where a high overlap indicated a high uncertainty, and vice versa. We had 29 dermatologists diagnose a series of skin lesions and rate their confidence. We compared these results to those of the network. The network achieved a sensitivity and specificity of 50% and 88%, comparable to the average dermatologist (respectively 68% and 73%). Higher confidence/less uncertainty was associated with better diagnostic performance both in the neural network and in dermatologists. We found no correlation between the uncertainty of the neural network and the confidence of dermatologists (R = −0.06, p = 0.77). Dermatologists should not blindly trust the output of a neural network, especially when its uncertainty is high. The addition of an uncertainty score may stimulate the human-computer interaction.

Abstract

Corticotropin-releasing hormone receptor-1 (CRH-R1) is expressed in human mast cells, but its role in skin diseases is unknown. By using a sequential double-staining technique, the mast cell expression of CRH-R1 was investigated in biopsies from lesional and non-lesional skin samples of patients with actinic keratosis (AK), basal cell carcinoma (BCC), squamous cell carcinoma (SCC) and psoriasis. Dermal tryptase⁺ mast cells expressed CRH-R1 immunoreactivity in the non-lesional skin in all patient groups. The CRH-R1 expression was significantly increased in the lesional skin of AK (p = 0.03) and psoriasis (p = 0.02), non-significantly in BCC (p = 0.129), but not increased in SCC. To investigate the regulation of CRH-R1, the LAD2 mast cell line was irradiated with UVB or stimulated with CRH or 1,25-dihydroxyvitamin D₃ [1,25-(OH)₂D₃]. Consequently, UVB at 90 mJ/cm² (p = 0.041) and 120 mJ/cm² (p = 0.039) decreased CRH-R1 expression. Instead, CRH at 100 and 1000 nM increased CRH-R1 immunostaining, but did not affect the proliferative response. The treatment with 10 and 100 nM 1,25-(OH)₂D₃ led to a noticeable increase in CRH-R1 staining. After irradiating with UVB, the concentration of CRH increased in the conditioned medium, but not in sonicated LAD2 mast cells. In conclusion, the lack of sufficient levels of CRH-R1 in mast cells may be related to diminished antitumoural response in SCC and possibly in BCC.

Abstract

Glucocorticoids (GCs) are widely used for the treatment of inflammatory skin diseases despite significant adverse effects including skin atrophy. Effects of GCs are mediated by the glucocorticoid receptor (GR), a well-known transcription factor. Previously, we discovered that one of the GR target genes, REDD1, is causatively involved in skin atrophy. Here, we investigated its role in GR function using HaCaT REDD1 knockout (KO) keratinocytes. We found large differences in transcriptome of REDD1 KO and control Cas9 cells in response to glucocorticoid fluocinolone acetonide (FA): both the scope and amplitude of response were significantly decreased in REDD1 KO. The status of REDD1 did not affect GR stability/degradation during self-desensitization, and major steps in GR activation—its nuclear import and phosphorylation at activating Ser211. However, the amount of GR phosphorylated at Ser226 that may play negative role in GR signalling, was increased in the nuclei of REDD1 KO cells. GR nuclear import and transcriptional activity also depend on the composition of GR chaperone complex: exchange of chaperone FKBP51 (FK506-binding protein 5) for FKBP52 (FK506-binding protein 4) being a necessary step in GR activation. We found the increased expression and abnormal nuclear translocation of FKBP51 in both untreated and FA-treated REDD1 KO cells. Overall, our results suggest the existence of a feed-forward loop in GR signalling mediated by its target gene REDD1, which has translational potential for the development of safer GR-targeted therapies.

Abstract

The risk of diabetes mellitus (DM) in vitiligo patients is higher than that in non-vitiligo population. Our goal was to explore the influencing factors for DM in vitiligo patients. A matched-pair design of 107 cases with DM and 428 controls without DM was conducted among vitiligo patients in Xijing hospital from January 2010 to October 2021. The baseline characteristics of patients were analysed based on standard descriptive statistics. The vitiligo-associated characteristics were analysed by logistic regression to identify influencing factors of DM. Interaction analysis was performed to explore the additive interactions between vitiligo-associated characteristics and baseline characteristics. After adjustment for the baseline characteristics, the severity of vitiligo [odds ratio (OR) = 2.47, 95% confidence interval (CI): 1.47–4.14] and onset age of vitiligo (OR = 0.98, 95% CI: 0.97–0.99) had a significant correlation with occurrence of DM. The severity of vitiligo had additive interaction with family history of diabetes [relative excess risk due to interaction (RERI) = 132.51 (95% CI: 5.51–1100.20), attributable proportion (AP) = 0.91 (95% CI: 0.17–0.95), synergy index (S) = 11.53 (95% CI: 1.32–100.5)] and with smoking history [RERI = 6.54 (95% CI: 0.67–19.83), AP = 0.64 (95% CI: 0.04–0.80), S = 3.48 (95% CI: 1.17–10.36)]. Earlier onset age of vitiligo and greater BSA involvement might be two independent risk factors for DM in vitiligo patients. Interaction assessment identified the severity of vitiligo as additive interaction factors with diabetes family history and with smoking history for the DM occurrence.

Abstract

Wound healing of deep burn injuries is often accompanied by severe scarring, such as hypertrophic scar (HTS) formation. In severe burn wounds, where the subcutis is also damaged, the scars adhere to structures underneath, resulting in stiffness of the scar and impaired motion. Over the recent years, a promising solution has emerged: autologous fat grafting, also known as lipofilling. Previous clinical reports have shown that the anti-fibrotic effect has been attributed to the presence of adipose-derived stromal cells (ADSC). In the proposed study, we aim to investigate the effect of fat grafting in 3D organotypic skin cultures mimicking an HTS-like environment. To this end, organotypic skin cultures were embedded with normal skin fibroblasts (NF) or HTS-derived fibroblasts with or without incorporation of human adipose subcutaneous tissue (ADT) and one part was thermally wounded to examine their effect on epithelialization. The developed skin cultures were analysed on morphology and protein level. Analysis revealed that ADT-containing organotypic skin cultures comprise an improved epidermal homeostasis, and a fully formed basement membrane, similar to native human skin (NHS). Furthermore, the addition of ADT significantly reduced myofibroblast presence, which indicates its anti-fibrotic effect. Finally, re-epithelialization measurements showed that ADT reduced re-epithelialization in skin cultures embedded with NFs, whereas HTS-fibroblast-embedded skin cultures showed complete wound closure. In conclusion, we succeeded in developing a 3D organotypic HTS-skin model incorporated with subcutaneous tissue that allows further investigation on the molecular mechanism of fat grafting.

Abstract

Interleukin (IL)-23–independent IL-17A production has been suggested to be involved in persistent manifestations of psoriatic disease, including anti–IL-12/23–refractory psoriatic plaques; this study aimed to test this hypothesis by investigating the clinical and molecular effects of direct IL-17A (with secukinumab) versus selective IL-23 inhibition (with guselkumab) in patients with anti–IL-12/23 (ustekinumab)–refractory psoriatic plaques. A 16-week, randomized, open-label, parallel-group, Phase IIa study (ARROW, NCT03553823) was conducted in patients with ≥1 active psoriatic plaque (total clinical score [TCS] ≥6) at screening despite treatment with ustekinumab, and a Psoriasis Area and Severity Index (PASI) score 1–10. Patients were randomized 1:1 to receive secukinumab 300 mg (n = 20) or guselkumab 100 mg (n = 20). Biopsies from one refractory (‘target plaque’) were obtained at baseline and Week 16. The primary endpoint was the proportion of patients whose ustekinumab-refractory target plaque achieved clear/almost clear status (TCS 0–2) at Week 16. Transcriptomic and histological analyses were conducted on target plaques to determine the molecular effects of direct IL-17A versus selective IL-23 inhibition. At Week 16, target plaque clear/almost clear status was achieved in 60.0% of patients treated with secukinumab versus 40.0% of patients treated with guselkumab (p = 0.1715). Molecular analyses identified that secukinumab modulated a greater proportion of psoriasis disease transcriptome genes (72.1% vs. 48.0%) and resulted in more histological responders (72.2% vs. 53.3%) compared with guselkumab. Secukinumab demonstrated a greater clinical and molecular effect on ustekinumab-refractory psoriatic plaques versus guselkumab. These results are consistent with the hypothesis that IL-23–independent IL-17 mechanisms may be relevant to the inflammation driving refractory manifestations of psoriasis.

Abstract

The molecular mechanisms underlying melanoma metastasis remain poorly understood. In this study, we aimed to delineate the mechanisms underlying gene expression alterations during metastatic potential acquisition and characterize the metastatic subclones within primary cell lines. We performed single-cell RNA sequencing of a poorly metastatic melanoma cell line (WM239A) and its subclones with high metastatic potential to the lung (113/6-4L) and the brain (131/4-5B1 and 131/4-5B2). Unsupervised clustering of 8173 melanoma cells identified three distinct clusters according to cell type (‘Primary’, ‘Lung’ and ‘Brain’ clusters) with differential expression of MITF and AXL pathways and putative cancer and cell cycle drivers, with the lung cluster expressing intermediate but distinct gene profiles between primary and brain clusters. Principal component (PC) analysis revealed that PC2 (the second PC), which was positively associated with MITF expression and negatively with AXL pathways, primarily segregated cell types, in addition to PC1 of the cell cycle pathway. Pseudotime trajectory and RNA velocity analyses suggested the existence of cellular subsets with metastatic potential in the Primary cluster and an association between PC2 signature alteration and metastasis potential acquisition. Analysis of The Cancer Genome Atlas melanoma samples by clustering into PC2-high and -low clusters by quartiles of PC2 signature expression revealed that the PC2-high cluster was an independent significant factor for poor prognosis (p-value = 0.003) with distinct genomic and transcriptomic characteristics, compared to the PC2-low cluster. In conclusion, we identified signatures of melanoma metastasis with prognostic significance and putative pro-metastatic subclones within a primary cell line.

Abstract

The gut microbiome is increasingly recognized to alter cancer risk, progression and response to treatments such as immunotherapy, especially in cutaneous melanoma. However, whether the microbiome influences immune checkpoint inhibitor (ICI) immunotherapy response to non-melanoma skin cancer has not yet been defined. As squamous cell carcinomas (SCC) are in closest proximity to the skin microbiome, we hypothesized that the skin microbiome, which regulates cutaneous immunity, might affect SCC-associated anti-PD1 immunotherapy treatment response. We used ultraviolet radiation to induce SCC in SKH1 hairless mice. We then treated the mice with broad-band antibiotics to deplete the microbiome, followed by colonisation by candidate skin and gut bacteria or persistent antibiotic treatment, all in parallel with ICI treatment. We longitudinally monitored skin and gut microbiome dynamics by 16S rRNA gene sequencing and tumour burden by periodic tumour measurements and histologic assessment. Our study revealed that antibiotics-induced abrogation of the microbiome reduced the tumour burden, suggesting a functional role of the microbiome in non-melanoma skin cancer therapy response.

Abstract

Background

Healthcare professionals (HCPs) should strive to create the maximum value for their patients in which value is defined as the patient-relevant health outcomes achieved per costs made. However, currently it remains difficult to determine which outcomes matter to an individual psoriasis patient.

Objective

To define outcome profiles, or so called ‘patient value profiles’, within a cohort of psoriasis patients that can be translated to daily practice to increase value for the individual patient.

Methods

Hierarchical clustering on principal components (HCPC) was used to identify groups of patients sharing the same profile within an outcome ranking exercise. Once the clusters were defined, their characterization was provided based on a V-test. In a final step, a multi-class decision tree (MDT) based on relevant socio-demographic and clinical variables was built to allocate patients to a cluster.

Results

In the ranking exercise 120 patients participated. The median age was 50.0 (IQR 25.0) years and 36.7% were female. Median PASI score was 2.4 (IQR 5.2) and median duration of psoriasis was 17.0 (IQR 20.0) years. Primary treatment varied from topicals to biologicals. We found three distinct patient value profiles in this cohort (QoL, cost and treatment). A MDT was built which had an accuracy of 64%.

Conclusion

We found three distinct patient value profiles in a cohort of psoriasis patients and patients can be easily assigned to one of these profiles based on a MDT. HCPs can use these profiles to steer psoriasis management accordingly allowing for a more goal-orientated approach.

Abstract

Melanoma is a notoriously radioresistant type of skin cancer. Elucidation of the specific mechanisms underlying radioresistance is necessary to improve the clinical efficacy of radiation therapy. To identify the key factors contributing to radioresistance, five melanoma cell lines were selected for study and genes that were upregulated in relatively radioresistant melanomas compared with radiosensitive melanoma cells determined via RNA sequencing technology. In particular, we focused on cyclin D1 (CCND1), a well known cell cycle regulatory molecule. In radiosensitive melanoma, overexpression of cyclin D1 reduced apoptosis. In radioresistant melanoma cell lines, suppression of cyclin D1 with a specific inhibitor or siRNA increased apoptosis and decreased cell proliferation in 2D and 3D spheroid cultures. In addition, we observed increased expression of γ-H2AX, a molecular marker of DNA damage, even at a later time after γ-irradiation, under conditions of inhibition of cyclin D1, with a response pattern similar to that of radiosensitive SK-Mel5. In the same context, expression and nuclear foci formation of RAD51, a key enzyme for homologous recombination (HR), were reduced upon inhibition of cyclin D1. Downregulation of RAD51 also reduced cell survival to irradiation. Overall, suppression of cyclin D1 expression or function led to reduced radiation-induced DNA damage response (DDR) and triggered cell death. Our collective findings indicate that the presence of increased cyclin D1 potentially contributes to the development of radioresistance through effects on RAD51 in melanoma and could therefore serve as a therapeutic target for improving the efficacy of radiation therapy.