After germinal center B cells undergo somatic mutation and antigen selection, they become either memory B cells or plasma cells, but the signal requirements that control entry into either pathway have been unclear. When purified human germinal center cells were cultured with interleukin-2, interleukin-10, and cells expressing CD40 ligand, cells with characteristics of memory B cells were generated. Removal of CD40 ligand from the system resulted in terminal differentiation of germinal center B cells into cells with the characteristics of plasma cells. These results indicate that CD40 ligand directs the differentiation of germinal center B cells toward memory B cells rather than toward plasma cells.

Human memory B cells that carry mutated IgV region genes were isolated from tonsils by negative selection of IgD+ naive B cells and CD38+ germinal center B cells and plasma cells. They were mainly found within the intraepithelial areas, but not in the B cell follicles of human tonsils. Memory B cells but not naive B cells have the capacity to present antigen directly to T cells, owing to the constitutive expression of the accessory molecules B7-1/CD80 and B7-2/CD86. Signals through antigen receptors and CD40 antigen result in these two molecules being further up-regulated more rapidly and strongly on memory B cells than on naive B cells. The unique anatomical localization of memory B cells beneath the surface of mucosa, together with their strong APC capacity, may explain the well-known prompt and robust secondary antibody responses.

Using a set of surface markers including IgD and CD38, human tonsillar B cells were classified into discrete subpopulations. Molecular and functional analysis allowed us to identify: i) two sets of naive B cells (Bm1 and Bm2); ii) germinal center founder cells (Bm2'); iii) an obscure population of germinal center B cells, displaying a high load of somatic mutations in IgV genes, C mu to C delta switch and preferential Ig lambda light chain usage: these cells may represent the precursors of normal and malignant IgD-secreting plasma cells; iv) the centroblasts (Bm3) in which somatic mutation machinery is activated; v) the centrocytes (Bm4) in which isotype switch occurs; vi) the memory B cells. The characterization of these subpopulations showed that: i) programmed cell death is set before somatic mutations, possibly providing an efficient way for affinity maturation; ii) only high affinity centrocytes are allowed to switch isotype; iii) CD40-ligation inhibits plasmacytic differentiation of mature B lymphocytes; iv) memory B cells preferentially differentiate into plasma cells; v) IgD isotype switch occurs in normal B cells; vi) receptor editing may be induced by somatic mutations in germinal centers. We also characterized two types of antigen-presenting cells in germinal centers: follicular dendritic cells that select high affinity B cells, and a new subset of germinal center dendritic cells that activate germinal center T cells.

Isolation of large numbers of surface IgD+CD38− naive and surface IgD−CD38− memory B cells allowed us to study the intrinsic differences between these two populations. Upon in vitro culture with IL-2 and IL-10, human CD40–activated memory B cells undergo terminal differentiation into plasma cells more readily than do naive B cells, as they give rise to five- to eightfold more plasma cells and three- to fourfold more secreted immunoglobulins. By contrast, naive B cells give rise to a larger number of nondifferentiated B blasts. Saturating concentrations of CD40 ligand, which fully inhibit naive B cell differentiation, only partially affect that of memory B cells. The propensity of memory B cells to undergo terminal plasma cell differentiation may explain the extensive extra follicular plasma cell reaction and the limited germinal center reaction observed in vivo after secondary immunizations, which contrast with primary responses in carrier-primed animals. This unique feature of memory B cells may confer two important capacities to the immune system: (a) the rapid generation of a large number of effector cells to efficiently eliminate the pathogens; and (b) the prevention of the overexpansion and chronic accumulation of one particular memory B cell clone that would freeze the available peripheral repertoire.

Human myeloma are incurable hematologic cancers of immunoglobulin-secreting plasma cells in bone marrow. Although malignant plasma cells can be almost eradicated from the patient's bone marrow by chemotherapy, drug-resistant myeloma precursor cells persist in an apparently cryptic compartment. Controversy exists as to whether myeloma precursor cells are hematopoietic stem cells, pre–B cells, germinal center (GC) B cells, circulating memory cells, or plasma blasts. This situation reflects what has been a general problem in cancer research for years: how to compare a tumor with its normal counterpart. Although several studies have demonstrated somatically mutated immunoglobulin variable region genes in multiple myeloma, it is unclear if myeloma cells are derived from GCs or post-GC memory B cells. Immunoglobulin (Ig)D-secreting myeloma have two unique immunoglobulin features, including a biased λ light chain expression and a Cμ–Cδ isotype switch. Using surface markers, we have previously isolated a population of surface IgM−IgD+CD38+ GC B cells that carry the most impressive somatic mutation in their IgV genes. Here we show that this population of GC B cells displays the two molecular features of IgD-secreting myeloma cells: a biased λ light chain expression and a Cμ–Cδ isotype switch. The demonstration of these peculiar GC B cells to differentiate into IgD-secreting plasma cells but not memory B cells both in vivo and in vitro suggests that IgD-secreting plasma and myeloma cells are derived from GCs.