Diagnostic 2-Gene Classifier in Early-Stage Mycosis Fungoides: A Retrospective Multicenter Study

Introduction

Mycosis fungoides (MF) is the most common type of cutaneous T-cell lymphoma and represents more than 50% of all primary cutaneous lymphomas. It is typically an indolent disorder with limited patches and plaques. One third, however, experience progression with ulcerating tumors and possible further systemic dissemination. Currently, the diagnosis of MF is based on clinical and histological examinations, but this has proven, especially in early-stage disease, to be challenging due to similarities with several benign skin conditions such as psoriasis, pityriasis lichenoides chronica, and dermatitis. Despite intensive research, reliable diagnostic biomarkers for early-stage MF are still needed. This study aims to identify a diagnostic classifier that could support the diagnostic workup leading to an exact diagnosis in the early-stage of this complex and potentially lethal disease.

Methods

We analyzed 43 formalin-fixed and paraffin-embedded (FFPE) skin biopsies from 36 patients with early-stage MF. Seven patients had 2 longitudinal biopsies available for analysis. These were compared with FFPE skin biopsies from patients with unspecified dermatitis (n=29) and healthy skin (n=12). All samples were collected from the archives of the Department of Pathology, Region Zealand, Denmark in the period 1990-2016. The histological samples were revised and clinical records were reviewed to establish the diagnosis and stage for each patient. Total RNA was extracted from ten 10-μm sections of FFPE tissue, and 50-100 ng of RNA was analyzed on the NanoString nCounter platform by applying the Myeloid Innate Immunity Panel, which quantifies the expression of 800 immune related genes. Differentially expressed genes (DEG, 2-fold change, p<0.05), were assessed by ANOVA, and a Support Vector Machine (SVM) diagnostic classifier was built based on 2-10 DEG and evaluated by 10-fold cross-validation. The classifier was tested on an independent, early-stage MF patient cohort (n=27). Protein expression was validated with immunohistochemistry and digitally analyzed by applying a specifically designed algorithm with the Leica Tissue IA 2.0 software. Double immunofluorescence staining protocols were developed to identify subtypes of TRAF1 positive cells in combination with various macrophage/dendritic cell markers (CD168, CD63, CD11c, CD1a, CD14, and S100) as well as T-cell (CD3) and B-cell (CD20) markers.

Results

A diagnostic classifier consisting of TOX and TRAF1 was able to distinguish early-stage MF from dermatitis with an overall accuracy of 85% in the discovery cohort and 80% in an independent validation cohort. TOX and TRAF1 protein levels were significantly elevated in early-stage MF compared to the dermatitis group (p < 0.0001). TOX and TRAF1 were also significantly increased in the progression from early-stage MF to tumor stage MF (p=0.003 and p=0.004, respectively). Subtypes of TRAF1-positive dendritic cells in early-stage MF consisted primarily of S100+ cells in both the epidermal and dermal compartment. A few TRAF1+ cells in the Pautrier microabscesses stained double positive with CD11c. In tumor stage MF the majority of TRAF1+ dendritic cells counterstained with CD1a and CD11c. Both neoplastic and reactive T-cells (CD3+) expressed TRAF1 in a minor degree in early-stage MF, while T-cells in tumor stage MF expressed TRAF1 in a much higher degree and the majority of the neoplastic T-cells were TRAF1 positive. No macrophages (CD68+ or CD163+) or B-cells double stained with TRAF1.

Conclusion

In the present study, we developed a two gene mRNA diagnostic classifier discriminating early-stage MF from dermatitis. The protein expression level of TOX and TRAF1 confirmed our gene expression levels and identified a highly significant difference between early-stage MF and dermatitis, which can prove useful in diagnostics of early-stage MF.

View largeDownload slide

View largeDownload slide

Close modal