Origin and pathogenesis of nodular lymphocyte-predominant Hodgkin lymphoma as revealed by global gene expression analysis

Abstract

The pathogenesis of nodular lymphocyte-predominant Hodgkin lymphoma (NLPHL) and its relationship to other lymphomas are largely unknown. This is partly because of the technical challenge of analyzing its rare neoplastic lymphocytic and histiocytic (L&H) cells, which are dispersed in an abundant nonneoplastic cellular microenvironment. We performed a genome-wide expression study of microdissected L&H lymphoma cells in comparison to normal and other malignant B cells that indicated a relationship of L&H cells to and/or that they originate from germinal center B cells at the transition to memory B cells. L&H cells show a surprisingly high similarity to the tumor cells of T cell-rich B cell lymphoma and classical Hodgkin lymphoma, a partial loss of their B cell phenotype, and deregulation of many apoptosis regulators and putative oncogenes. Importantly, L&H cells are characterized by constitutive nuclear factor kappaB activity and aberrant extracellular signal-regulated kinase signaling. Thus, these findings shed new light on the nature of L&H cells, reveal several novel pathogenetic mechanisms in NLPHL, and may help in differential diagnosis and lead to novel therapeutic strategies.

Figure 1.

Unsupervised hierarchical clustering of lymphoma samples. The dendrogram is based on the expression data of the 582 most variable genes (BL, black; cHL, red; DLBCL, dark blue; FL, light blue; NLPHL, orange; TCRBL, green).

Figure 2.

Relatedness of L&H cells with other lymphomas and with normal B cells by PCA. (A) 3,364 genes, whose median expression was above background in at least one lymphoma entity, were used by PCA to assess the relatedness of L&H cells (mean of 5 samples) to the malignant B cells of other lymphomas (mean of 4–12 samples). The first two principal components are shown (accounting for 88.1% and 5.5% of the total variance, respectively). (B) Genes discriminating between GC B cells and naive B (N), memory B (M), or plasma cells (PC; FC ≥ 4; FDR < 0.05 after t test; 229, 128, and 130 probe sets), and between naive and memory B cells (FC ≥ 2.5; FDR < 0.05 after t test; 47 probe sets) or between memory B cells and plasma cells (FC ≥ 4; FDR < 0.05 after t test; 135 probe sets) were used by PCA to identify the closest normal counterpart of L&H cells. In each of the five graphs, the y axis indicates the first principal component score, the horizontal solid lines show the median score of each cell subset, and the horizontal dotted line represents the mean of the median scores of the two normal subsets being compared. Also shown are the p-values (obtained by t test) of the pairwise comparisons (vertical lines) between L&H cells and normal B cells. The percentage of total variance accounted for by the first principal component is as follows: GC B versus plasma cells, 89%; GC B versus naïve B cells, 93.9%; GC B versus memory B cells, 92.5%; memory versus naïve B cells, 86.6%; and memory B versus plasma cells, 92.8%.

Figure 3.

Partial loss of B cell markers in NLPHL. (A) Heat map of the absolute expression values of B cell markers, which are encoded by the color bar (base 2 logarithmic scale). (B) Individual samples (squares) are placed along the first principal component (which accounts for 92.2% of the total variance) of the expression matrix of 61 selected B cell markers. Box plots summarize the score of each group in this component. The vertical lines ending in horizontal lines (whiskers) extend to the most extreme data points if they extend not more than 1.5 times the box width (the interquartile range). Otherwise, the whiskers end at an observed value that is not more than 1.5 times the box width away from the box. p-values of the pairwise comparisons using a Wilcoxon test are indicated.

Figure 4.

NF-κB activity in NLPHL. The list of NF-κB target genes experimentally validated in cHL cell lines (reference 89) was expanded by including other NF-κB target genes described in the literature or other sources (available at http://www.nf-kb.org). After excluding all noninformative probe sets (signals below background in all samples), we obtained a list of 62 probe sets, corresponding to 50 NF-κB target genes, of which 48 turned out to be consistently up-regulated in L&H versus GC B cells (FDR < 0.05 after t test; FC ≥ 1.8). (A) Heat map of NF-κB target genes. The sources of the NF-κB target genes are (1) http://www.nf-kb.org, (2) the literature, and (3) Hinz et al. (reference 89), as indicated in parentheses after the gene names. The absolute expression values on a logarithmic scale (base 2) are encoded by the color bar. (B) Individual samples (squares) are placed along the first principal component (which accounts for 80.1% of the total variance) of the expression matrix of 50 NF-κB target genes. Box plots summarize the scores displayed by each group in this component. The vertical lines ending in horizontal lines (whiskers) extend to the most extreme data points if they extend not more than 1.5 times the box width (the interquartile range). Otherwise, the whiskers end at an observed value that is not more than 1.5 times the box width away from the box. p-values of the pairwise comparisons using a Wilcoxon test are indicated.

Figure 5.

Protein expression in NLPHL cases. (A–J) Immunohistological stainings of paraffin sections of lymph nodes from patients with NLPHL for active NF-κB p65 subunit (A), total ERK (B), active p-ERK (C), CTSB (D), MMP9 (E), MMP12 (F), LGALS3 (G), IL21R (H), EOMES (I), and p-STAT1 (J). The anti–NF-κB p65 antibody recognizes only p65 that is no longer bound by IκB factors, and thus identifies active NF-κB factors, although the staining appears often predominantly cytoplasmatic (reference 45). Insets show L&H cells at higher magnification. Some insets (A, B, D, E, G, and J) are magnified using Photoshop CS software (Adobe). Bars, 10 μm.