Transcriptional atlas of normal PC development and neoplastic transformation. Data were normalized with RUVSeq, and differential expression of genes was analyzed with Deseq2 or masigpro, followed by k-means clustering in R. (A) Tridimensional principal component analysis of gene-expression data based on the top 500 variable genes. Normal PCs were isolated by fluorescence-activated cell sorting (FACS) from SLOs (ie, tonsils, SLO) and PB according to immunoglobulin heavy-chain isotypes, as well as from BM samples of healthy adults according to differential expression of CD19, CD39, CD56, and CD81. Tumor PCs were isolated by FACS from BM samples of patients with AL (n = 32), MGUS (n = 6), and MM (n = 32). (B) Heat map of gene expression in normal PC subsets (after merging SLO- and PB-PCs with immunoglobulin M [IgM], IgG, and IgA isotypes as well as all CD39⁻ BM-PCs due to overlapping profiles) and tumor PCs from AL, MGUS, and MM patients, based on a total of 2200 genes selected by their differential expression in normal SLO-PCs, PB-PCs, and CD39⁺ and CD39⁻ BM-PCs plus 720 genes differentially expressed in tumor PCs from AL, MGUS, and MM patients. TPs were defined by semisupervised k-means clustering (k = 13), and gene expression is represented with a row z score. (C) Functional enrichment heat map based on top nonredundant biological functions (ie, gene ontology [GO]) (P < .05) per TP determined with Metascape. TP1, TP7, and TP8 shared the same top biological functions and were grouped. Color intensity is proportional to the significance level. Box plots from left to right represent gene-expression levels in normal SLO-PCs, PB-PCs, and CD39⁺ and CD39⁻ BM-PCs as well as tumor PCs from AL, MGUS, and MM patients. The most significant deregulated genes are shown at the right end of the graphical representation. (D) Uniform manifold approximation and projection (UMAP) of 142 778 individual PCs isolated by FACS from healthy adults; SLOs (n = 3), PB (n = 3), and BM (n = 3) and from patients with AL (n = 6) and MM (n = 15). (E) Single normal BM-PCs and tumor PCs were colored according to the TPs identified in panel C, based on the expression level of the most significant genes defining each TP. We used the Seurat R package²⁴ for data integration and clustering.