Identification of human germinal center light and dark zone cells and their relationship to human B-cell lymphomas

Abstract

Germinal centers (GCs) are sites of B-cell clonal expansion, hypermutation, and selection. GCs are polarized into dark (DZ) and light zones (LZ), a distinction that is of key importance to GC selection. However, the difference between the B cells in each of these zones in humans remains unclear. We show that, as in mice, CXCR4 and CD83 can be used to distinguish human LZ and DZ cells. Using these markers, we show that LZ and DZ cells in mice and humans differ only in the expression of characteristic "activation" and "proliferation" programs, suggesting that these populations represent alternating states of a single-cell type rather than distinct differentiation stages. In addition, LZ/DZ transcriptional profiling shows that, with the exception of "molecular" Burkitt lymphomas, nearly all human B-cell malignancies closely resemble LZ cells, which has important implications for our understanding of the molecular programs of lymphomagenesis.

Figure 1

CXCR4 and CD83 define human LZ and DZ B-cell populations. (A) Flow cytometric profile of day 10 mouse lymph node GCs and human tonsil GCs stained for markers CXCR4 and CD83. LZ and DZ gates and percentages are shown. Gating as shown in supplemental Figure 1A. (B-D) Each plot is representative of at least 4 independent experiments. Immunofluorescent staining of frozen tonsil (B-C) or paraffin-embedded reactive lymph node (D) samples showing the anatomic distribution of CXCR4 (B), CD86 (C), and CD83 (D) in human GCs. Light zones are defined by CD23 staining (green). Note that CD23 and CD86 staining appears to overlap in the LZ because of insufficient resolution of confocal microscopy to discern molecules expressed on juxtaposed membranes. The GC perimeter is defined in frozen and paraffin-embedded sections by counterstainings with IgD and DAPI, respectively (both in blue). Scale bars represent 100 μm. Histology data are representative of 2 independent experiments. Image acquisition parameters are described in “Microscopy.”

Figure 2

Phenotype of LZ and DZ B cells in mice and humans. (A) Cell-cycle (DNA content) profiles of human and mouse LZ and DZ cells (gated shown in supplemental Figure 1A). Multiple experiments (n = 5 for mouse and n = 3 for human) are quantified in the graphs on the right (error bar represents SD). (B) Forward scatter (FSC) and side scatter (SSC) profiles of human and mouse LZ, DZ, and naive B cells. Naive B cells are defined as CD19⁺IgD⁺CD83⁻ in human tonsil and as B220⁺FAS⁻CD38⁺ in mouse lymph node. Right panels: quantification of multiple experiments (n = 4 for mouse and n = 3 for human; bar represents SD). (C) Expression of selected markers by human and mouse LZ, DZ, and naive B cells. (D) Expression of selected B-cell markers by human LZ, DZ, and naive B cells. (E) Reproducibility of surface molecule expression between human samples. Graph represents quantification of multiple experiments (n = 2 for CD86, MHC class II; n = 3 for Igκ+λ; and n = 4 for the remaining markers; bar represents SD).

Figure 3

Gene expression in mouse and human LZ/DZ. (A) Unsupervised clustering of microarray data from LZ and DZ cells sorted from mouse lymph node or human tonsil. (B) Volcano plots showing global differences between naive and GC, plasma cell and GC, and LZ and DZ B cells in human and mouse. Microarray data for naive and total GC B cells and plasma cells data were obtained from Luckey et al and Long et al. Green lines indicate 2-fold differences (x-axis) and P = .05 (y-axis). Numbers in the plots indicate the number of probes differing by 2-fold or more (P < .05 in each of the comparisons). The total number of probes plotted for each species is indicated on the left.

Figure 4

Polarization of gene expression in LZ and DZ B cells. (A) Heat maps showing differential expression in LZ and DZ of selected genes in mice and humans. Colors indicate fold change (Log₁₀ base) between one zone and the opposite zone within the same sample (human) or pool (mouse). (B) Immunofluorescence of human (top) and wild-type mouse (middle) GCs showing the anatomic distribution of AID protein. AID is stained in both species using the same rat monoclonal antibody (mAID-2). Light zones are defined by CD23 staining (green) in humans and Ig (anti-IgG, heavy and light chains, which stains mostly immune complexes deposited on follicular dendritic cells) in mice. The GC perimeter is defined by counterstaining with DAPI in human and Bcl-6 in mouse (both in blue). Immunofluorescent staining of an Aicda^−/− lymph node is shown as a control for the specificity of the anti-AID antibody (bottom). Scale bars represent 100 μm. Image acquisition parameters are described in “Microscopy.” (C) Mouse LZ and DZ signatures overlaid on human gene expression data, plotted as expression scatter plots (left) or volcano plots (right). Human gene expression data are shown (gray) with genes contained in mouse LZ or DZ signatures highlighted in blue or red, respectively. Interspecies orthologs defined as genes bearing the same Official Gene Symbol. Volcano plots: dotted line represents 2-fold P < .05. (D) Human LZ and DZ signatures overlaid on mouse gene expression data. Details are as in panel C.

Figure 5

Most GC-derived B-NHLs share an LZ-related phenotype. (A) Class prediction of GC-derived B-NHLs using the Weighted Voting algorithm. The dataset reported by Caron et al, where tonsillar GC B-cells were isolated according to their surface expression of CXCR4, was used as a validation set. (B) Class prediction in the same B-NHL samples using the SPLASH algorithm within the Bluegenes tool. In this analysis, DLBCL cell lines were included. All DLBCL primary cases and cell lines are coded based on their cell-of-origin classification (ABC/GCB). (C) Hierarchical clustering of the same GC-derived B-NHL based on the expression of the “compound pathway signature,” as described in the main text. Note the presence of 2 main clusters (DZ-like, LZ-like). Red represents highly expressed genes; and blue, lower-expressed genes. The behavior of the signature in normal GC LZ/DZ B-cells is plotted in the accompanying heat map (left panel). DLBCL-PT indicates primary DLBCL cases; and DLBCL-CL, DLBCL cell lines.

Figure 6

LZ/DZ-related pathways underlie the distinction between molecular subgroups in aggressive mature B-NHL. (A) Consensus clustering (1000 bootstraps) of the aggressive mature B-NHL case series described by Hummel et al, according to the expression pattern of the common human/mouse signature (1.5-fold cutoff; supplemental Figure 4; supplemental Table 4). Shown is the clustering image with k = 2 (no significant improvement of the CDF was observed with higher k values). (B) Distribution of mBL and non-mBL cases, as defined by Hummel et al, among the 2 different subgroups identified by the consensus clustering analysis shown in panel A. The P value shown refers to the significance of the distribution of the 3 subclasses in Hummel et al (mBL, intermediate, non-mBL) among each class (DZ-like or LZ-like; χ² test). (C) GSEA plots illustrating the enrichment for mBL and non-mBL classifier gene signatures (as defined by Hummel et al) in human LZ and DZ GC B-cell gene expression data. Nominal and adjusted P values and false discovery rate are all below the detection level (< .001) for both comparisons.