Transcriptome in paraffin samples for the diagnosis and prognosis of adrenocortical carcinoma

Figures

• 

  Figure 1. FFPE transcriptome classification of ACT in the training cohort. (A) Unsupervised classification of ACT based on the 2000 most variable genes identifies three main groups, corresponding to ‘C1A’ aggressive carcinomas, ‘C1B’ indolent carcinomas and ‘C2’ adenomas. (B) Heatmap of the three non-negative matrix factorization (NMF) ranks used for generating the unsupervised classification. (C) Expression of steroid (adrenal differentiation), proliferation and immune signatures in each ACT. (D) Expression profiles of the top 50 most significantly different genes among the three transcriptome groups. (E) Consensus matrix representing the similarity between samples over iterative clustering algorithms. Rows and columns are patient samples. Consensus matrix values range from 0 (never clustered together) to 1 (always clustered together).

• 

  Figure 2. FFPE transcriptome classification in different tumor regions. (A) Hematoxylin/eosin/safran staining (40× magnification) of the two different tumor regions sampled for five patients in the training cohort, with one aggressive (region a) and one less aggressive (region b) region. The black bar represents 50 µm. Black arrows indicate mitosis. White arrows indicate atypical mitoses. (B) Description of the two different tumor regions sampled for the five patients with one aggressive (-a) and one less aggressive (-b) region. N/A: Fuhrman nuclear grade is not applicable for oncocytic cells. (C) Samples projection based on the two principle components (Dim1 and Dim2) of the PCA performed based on the 2000 most variable genes in the 95 unique patients of the training cohort. Samples from the training cohort are presented as faint circles colored by transcriptome class. Samples from two different tumor regions for eight patients are presented as full triangles colored by patient.

• 

  Figure 3. FFPE transcriptome classification and prognosis in the ACC validation cohort. (A) Samples projection based on the two principle components (Dim1 and Dim2) of the PCA performed based on the 2000 most variable genes in the 95 unique patients of the training cohort. Samples from the training cohort are presented as faint circles colored by transcriptome class. Samples from validation cohort are presented as full triangles colored by transcriptome class. (B) Disease-free survival according to the predicted transcriptome class in stage I–III ACC patients of the validation cohort. (C) Overall survival according to the predicted transcriptome class in stage I–IV ACC patients of the validation cohort.