Divide and conquer: the importance of cell division in regulating B-cell responses


Dr Stuart Tangye, Centenary Institute of Cancer Medicine and Cell Biology, Locked Bag #6, Newtown 2042, NSW, Australia. E-mail: s.tangye@centenary.usyd.edu.au


Proliferation is an essential characteristic of clonal selection and is required for the expansion of antigen reactive clones leading to the development of antibody of different isotypes and memory cells. New data for mouse and human B cells point to an important role for division in regulating isotype class and in optimizing development of protective immunity by the regulated entry of cells to the plasma cell lineage.


B-cell differentiation is classically thought of as a decision tree, where a sequence of signals is needed first for progression of the resting B cell to an activated state, followed by isotype switching, and then for development either to memory cells or to short-lived or long-lived plasma cells (PCs). This regulated branching view is consistent with the discovery of a large number of cytokine signals that affect B-cell responses, and the cellular complexity of microenvironments such as germinal centres (GCs). This heterogeneity seems sufficient to account for the remarkably rich variation in cell phenotype that can be observed in vivo. Nevertheless, it is curious that a great deal of the B-cell response leading to immunoglobulin (Ig) production can be reproduced in vitro under highly controlled conditions. It is while exploring these latter systems that additional intrinsic components of the B-cell ‘decision tree’ emerge and indicate an important role for division itself in regulating the likelihood of various cell fates (summarized in Fig. 1a). These studies have provided new insights into B-cell biology, and the regulation of the production of Ig.

Figure 1.

The division hypothesis and stochastic division-linked regulation of isotype switching, antibody secretion and memory. (a) Resting IgM+ B cells are activated to proliferate following exposure to antigen (Ag). The combination of signals received, as well as intrinsic internal ‘programming’, dictates the progressive change in the probability of switching to a new isotype per division. Switched cells are shown as the blue cells at the top of the diagram. Both switched and unswitched cells undergo differentiation to immunoglobulin (Ig)-secreting cells (ISCs) at a similar rate, giving rise to Ig. When Ag stimulation ceases, the pool of activated cells undergoes contraction, as a result of apoptosis, leaving a core of switched and unswitched memory cells as well as long-lived ISCs (plasma cells). Upon restimulation, memory cells differentiate to ISCs more rapidly per division than naïve cells. (b) The top panel shows how the rate of switching can be represented as a probability map showing the proportion of undifferentiated cells that will change between each division cycle. The relation between arrow strength in (a) and probability is given for illustration. The middle panel shows how alternative differentiation events can have a different relation to division number. IgG1 switching in response to interleukin (IL)-4 proceeds earlier than IgE switching. The bottom panel shows how the relation to division can be modified by changing the cytokine concentration. (c) The bottom panel shows a putative set of overlapping differentiation maps associated with the combination of stimuli received during B-cell activation. The two top panels show how cells dividing according to this scheme will automatically generate considerable heterogeneity. The colour of the cells corresponds to the colour of the symbols on the map. The top panels also illustrate how two different rates of Ag persistence can drive the cells to different patterns and mixtures of cell types. After a period of contraction, the isotype mix of the memory (Mem) cells and long-lived ISCs will be different.

Isotype switching by murine and human naïve b cells

The importance of cytokines in inducing Ig isotype switching is well accepted. Interleukin (IL)-4, interferon (IFN)-γ and transforming growth factor (TGF)-β in mice1–8 and IL-4, IL-10, IL-13, IL-21 and TGF-β in humans9–17 have been shown to promote an increase in (a) the frequency of switched cells, (b) the release of switched Ig, and (c) the induction of molecular intermediates associated with switching. Nevertheless, B cells stimulated in the presence of these powerful ‘switch’ factors do not all switch. For example, after 3–10 days in culture, ∼30% of murine B cells will become IgG1+ in response to IL-4, ∼5% will switch to IgE and only 1–2% will switch to express IgA in the presence of TGF-β.18–21 When human B cells were examined, a similarly low frequency of Ig isotype switching at the cell surface level was observed.12,22–24 These results clearly revealed that other variables were involved in determining the efficacy of switching.

The discovery of an important missing variable arose from analysis of B-cell responses using the 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE) division tracking dye method introduced by Lyons and Parish.25 When B cells were stimulated in vitro with T-cell dependent (TD) or T-cell independent (TI) stimuli, they could be found distributed around a broad range of division numbers after a few days in culture, indicating significant variation in proliferation times.26,27 This variation enabled the detection of cells that had undergone a range of divisions at each single time of cell harvest. Staining for surface IgG1 showed that cells in early divisions (0–3) very rarely switched, irrespective of the time spent in culture.26,27 In contrast, after the third division, the frequency of cells that had switched to IgG1 increased with each division, reaching a peak at around 50% by divisions 7–8. It was also striking that the proportion of switched cells in each division number was constant on consecutive days, even though there had been significant progression in the average division reached by the cell population.26–29 Thus, division number itself, rather than time spent in culture, appeared to be the more important variable in regulating switch frequency. Other markers, such as syndecan-1 (CD138), found on Ig-secreting cells (ISCs)30 also appeared only in the later divisions.26,31 This phenomenon implied that progression through the division cycle itself may play a role in altering the likelihood or probability of switching. When a differentiation event is found to vary in closer relation to division number than to time in culture, we refer to the control mechanism as being ‘division-linked’. This division-linked relationship was found for all mouse Ig isotypes induced under different stimulation and cytokine combinations.26–28,32 In some cases, a cytokine was not involved. For example, induction of Ig isotype switching to IgG3 by stimulation with lipopolysaccharide (LPS) alone was division-linked.28 Furthermore, not all Ig isotypes displayed the same relation with division. Specifically, although culture with CD40L and IL-4 induced both IgG1 and IgE, the latter isotype required more divisions than the former.27 Varying the concentration of cytokines had the interesting effect of altering the relationship with division, with cells taking progressively more divisions to switch as the concentration of the ‘switch factor’ was lowered.27 Thus, the requirement for progression through division accounted for, to a large extent, the paradox of why powerful switching signals induced only a proportion of switched cells. First, within a stimulated population there is considerable heterogeneity in the number of divisions the cells have undergone, and, secondly, each cytokine and concentration results in a different pattern of switching with division number. Some cytokines exert additional effects on B-cell behaviour: IL-4 simultaneously promotes survival, proliferation and Ig isotype switching, thus yielding a dramatic amplification in the number of switched cells.7,26,27,33–36 In contrast, TGF-β is a switch factor for IgA4,6 while being a potent inhibitor of cell proliferation.9,37 Thus, even though those cells that reach later divisions can switch at high frequency, only very few achieve this number of divisions, accounting for the low overall efficiency of switching observed at the population level.28 This fundamental relationship between division and switching was also observed for human cells, with IL-4 and IL-13, and to a lesser extent IL-10, inducing Ig isotype switching of naïve B cells to IgG, predominantly IgG1.29 In contrast to mouse B cells, these cytokines did not seem to qualitatively alter the range of isotypes expressed by human B cells, as IgG1, IgG2 and IgG3 could all be detected, albeit at different frequencies, within the same divisions.29


Not all cells switch at the same division; rather the pattern of switching can approximate a normal distribution around a mean division number26–29,32 (summarized in Fig. 1b). Furthermore, not all cells will switch even after numerous divisions. Are these differences attributable to heterogeneity in the starting population with cells committed to one or other isotype, or are all cells multipotential for switching and stochastic internal events give rise to the final outcome? Compelling evidence in support of the latter perspective was obtained from studying two intersecting isotype switching patterns.28 In these experiments, LPS induced IgG3 from division 3; adding TGF-β induced switching to IgG2b over similar divisions and a small reduction in IgG3.28 Thus, in the presence of TGF-β, two division-linked switching events are proceeding within the activated B cells. If they were different subpopulations then little intersection would be observed. However, if the two were independent ‘stochastic’ processes operating in each cell then the net outcome could be predicted by knowing the outcome of each response. IgG2b switching must be dominant as it deletes the chromosomal region for IgG3.2,9 Careful measurements and calculation closely predicted the small amount of IgG3+ cells that were lost from each division.28 Furthermore, sorting IgG3 B cells and de novo-generated IgG3+ B cells present in the same culture revealed that each population was equally likely to undergo Ig isotype switching to yield IgG2b+ cells, consistent with separate independent stochastic mechanisms and inconsistent with separate switch committed precursors.28 These experiments helped to develop the concept of independent probability maps of differentiation events associated with division. Knowledge of these maps enables the prediction of the emergence of mixtures of differentiated cells over time (Figs 1b and c).27–29,32

The division hypothesis

These observations linking the acquisition of different isotypes with division suggested a novel scheme for regulation of immune class. In its simplest form, B cells have an internally encoded differentiation machinery linked to cell division, and class switching will proceed in concert with clonal expansion. As antigen (Ag) is required to continue driving proliferation, a feedback would operate: if a successful class(es) of antibody is produced then the Ag is cleared and further cell division stops. Persistence of Ag, and therefore unsuccessful clearance, would continue to stimulate cells to divide and thus switch to later isotypes (see Fig. 1c). For example, the relatively late expression of IgE and IgA in division rounds even under optimal cytokine conditions suggests that continued exposure to Ag would be required to provoke cells to these isotypes, perhaps accounting for their relative rarity.27,28

If Ag persistence is required to maintain B-cell proliferation, it would be expected that withdrawing stimulation in vitro would stop proliferation. This prediction was addressed directly by monitoring the consequence of withdrawal of stimuli that would be provided by engaging a T helper (Th) cell. Activated mouse B cells continued to divide a few times and showed a progressive loss of viability after the stimulus (CD40L) was removed. Importantly, the division-linked increases in switching proceeded even without the continued presence of stimuli.38 Human B-cell blasts show even more marked cessation of proliferation upon removal of stimulus, essentially stopping immediately.39,40 Thus, B cells will have to periodically re-engage T-cell help to continue clonal expansion, consistent with the hypothesis that Ag clearance would limit Ig isotype switching.

We have used the division hypothesis as a useful quantitative framework to help explore a series of questions relating to B-cell differentiation. These include the basis for the quantitative differences in the proliferation and differentiation of naïve and memory cells, the role of signals in promoting development of ISCs with division, and the independence of differentiation events taking place within the same cells.

Naïve versus memory b cells

Immunological memory is characterized by a rapid and robust immune response following re-exposure to the original immunizing Ag.41,42 The differences in primary and secondary humoral immune responses can be attributed to intrinsic differences between naïve and memory B cells, which endow these cells with unique properties and features. First, naïve and memory B cells differ phenotypically. Memory B cells have traditionally been identified by the loss of expression of IgM and IgD, and the acquired expression of switched Ig isotypes, such as IgG, IgA and IgE.43–46 Additional differences in phenotype include increased expression of major histocompatibility complex (MHC) class II molecules, as well as the activation molecules CD80, CD86 and CD95.45,47,48 The different cytoplasmic domains of switched Ig isotypes compared to those of IgM and IgD have been reported to favour a more robust response by memory B cells.49,50 Similarly, increased expression of B7 molecules (CD80 and CD86), and presumably MHC class II, provides memory B cells with the ability to act as potent Ag-presenting cells for CD4+ T cells (ref.45 and unpublished data). Secondly, naïve and memory B cells exhibit distinct responses following exposure to an activating stimulus. For instance, in vivo studies demonstrated that Ag-binding/memory B cells differentiated into Ig-secreting effector cells more rapidly than naïve/non-Ag-binding B cells.47,49,51–53 Such in vivo studies have been complemented by in vitro analysis of B-cell subsets, which found that (a) the threshold for activation of memory B cells is lower than that of naïve B cells with respect to the amount of Ag and T-cell help required for the induction of an effector response,54 (b) IgD human memory B cells yield large numbers of ISCs or CD38+ plasma-like cells, while IgD+ naïve B cells remain relatively undifferentiated (i.e. CD38),55–57 and (c) Ag-specific IgG was produced by IgD human memory, but not IgD+ naïve, B cells in the presence of specific Ag.45,58,59

It is possible to study memory B-cell responses in humans because naïve and Ag-specific memory B cells can be resolved, and thus purified, from one another by the differential expression of CD27.60,61 CD27+ memory cells represent ∼25%61,62 and ∼50%60,63 of peripheral blood and splenic B cells, respectively, and comprise not only IgG+ and IgA+ B cells, but also a significant number (up to 50%) of IgM+ memory B cells.63,64 This suggested that some of the attributes ascribed to naïve B cells, often isolated as IgM+ B cells, may in fact reflect the behaviour of contaminating residual IgM-expressing memory B cells. It is also quite convenient, and advantageous, to examine human B cells not only because of the ability to isolate large numbers of polyclonal memory B cells, but also because of the direct relevance to human B-cell dyscrasias. Thus, we have examined more closely the differences in the behaviour of human naïve and memory B cells with respect to proliferation, differentiation and Ig secretion.


Analysis of proliferation of CFSE-labelled human B cells revealed that memory B cells underwent more rounds of cell division than naïve B cells in response to stimulation with either CD40L alone or IL-10.39,64,65 These findings were consistent with increased proliferation of memory versus naïve B cells as determined by incorporation of 3H-thymidine.64,66 One of the mechanisms found to contribute to the increased rate of proliferation of memory B cells was that they entered their first division on average ∼24 hr earlier than naïve B cells following stimulation with CD40L alone. The earlier entry of memory B cells into division was independent of the Ig isotype expressed, because the times to first division of IgM+ and IgG+/IgA+/IgE+ memory B cells were similar.64 Remarkably, despite this difference in the time to first division, the subsequent rates of proliferation of the naïve and memory B cells were similar.64 Human naïve and memory B cells were also found to exhibit profound proliferative differences when either stimulated with CpG oligonucleotides67,68 or cocultured with human natural killer (NK) T cells,65 as evidenced by an increased frequency of memory B cells undergoing more rounds of cell division compared with naïve B cells, again determined by dilution of CFSE. Although it has not been examined, it is likely that under these stimulatory conditions memory B cells are induced to enter division earlier than naïve B cells, thus accounting for their greater proliferative response.

Additional studies revealed that memory B cells respond to some stimuli that are incapable of eliciting a response from naïve B cells. For instance, while the addition of IL-10 increased the proliferation of both populations of CD40L-stimulated human B cells, further addition of IL-2 led to the selective expansion of memory B cells.39,64 Similarly, although memory B cells responded to CpG alone, naïve B cells only responded to CpG when costimulated through either the B cell receptor (BcR)68 or IL-15 receptor.67 These differences in sensitivities to exogenous stimuli are likely to reflect increased expression by memory B cells of IL-2R64,66 or the CpG receptor toll-like receptor (TLR)968 compared to naïve B cells. These data predict that the differentiation of naïve B cells into memory B cells in vivo is accompanied by a cellular re-programming process resulting in changes in the quality and quantity of the responsiveness of the resultant memory B cell. Thus, memory B cells possess an intrinsic advantage over naïve B cells which allows accelerated entry of a greater frequency of memory cells into cell division; this is likely to contribute to the rapid humoral immune response characteristic of a secondary exposure to an immunizing Ag.


Human B cells

One of the earliest demonstrations that the differentiation of human B cells into effector cells required cell division was reported more than 20 years ago by Lipsky's group.69 In this study, it was reported that differentiation of activated human peripheral blood B cells into ISCs was abrogated when cultures were performed in the presence of hydroxyurea or mitomycin C, which are inhibitors of DNA synthesis. Delayed addition of these inhibitory drugs to the cultures also prevented the generation of ISCs, indicating that the ability of B cells to develop into effector ISCs was acquired during the process of cellular activation and division. It was further revealed that the ISCs generated during these in vitro cultures were a population of rapidly dividing cells, and that continued proliferation was required to sustain Ig production by these cells.69 This elegant study has been recently revisited by examining the differentiation of CFSE-labelled human B cells into effector cells, measured either by the acquired expression of CD38 and loss of CD20 (i.e. assuming a CD38++ CD20± PC phenotype)39,64,67 or by the accumulation of functional ISCs.39,70 Collectively, these studies revealed that memory, but not naïve, B cells efficiently yield effector ISCs, and that the frequency of such cells generated in culture increases with cell division.39,64,67,70 The critical regulators of this process appear to be CD40L and IL-10,39,56,71–73 although the contribution of engagement of the BcR or TLRs cannot be discounted.67,68 The requirement of cell division for the generation of PCs from precursor cells was also recently demonstrated in vivo,74 thereby validating the in vitro use of culture models for examining B-cell differentiation.

It was also found that, during the differentiation process, the proliferative behaviour of the CD38+ cells, which contain ISCs, changed such that their rate of proliferation exceeded that of the CD38 B cells by ∼ 2-fold.39,64 Furthermore, the CD38+ B cells generated in the cultures became independent of CD40L for their continued proliferation and survival. In contrast, CD38 B-cell blasts that had undergone several rounds of division remained highly dependent on the presence of CD40L for their continued survival; withdrawal of CD40L resulted in apoptosis of > 80% of these cells.39,40,64,71,72 These studies demonstrated that ISCs derived in vitro no longer required T-cell help for their survival, and suggested that they may correspond to plasmablasts detected in vivo in both humans75 and mice.76,77 These studies confirmed earlier reports that (a) ISCs generated in vitro acquire an increased rate of proliferation,69 and (b) following initial activation and expansion of human B cells by T-cell help, the resulting effector B cells can survive and function without the provision of activated T cells.78

A contentious issue in this area has been the role of CD40L in the fate of differentiating B cells. Several studies have reported that sustained engagement of CD40 on activated B cells promotes their development along a ‘memory’ pathway, while concomitantly inhibiting the generation of ISCs.56,71,72,79 The conclusion of these studies was that T-cell contact, in particular the provision of CD40L, maintained memory cell proliferation, whereas upon loss of contact with the activated T cell, presumably due to Ag clearance, the memory cell would differentiate into a PC. These results have been difficult to reconcile with other studies showing that simultaneous engagement of the BcR and CD4045,58 or coculture with T cells80 is required for the production of Ag-specific Ig, and that sustained culture of activated B cells with either activated T cells78 or CD40L and IL-1072,73 results in much higher numbers of ISCs, as well as higher levels of secreted Ig than obtained when cells are cultured without T cells or CD40L. The CD40L ‘loss’ explanation of the memory/PC decision is also difficult to reconcile with the proposal that memory cells can act as ‘stem cells’ for the expansion and maintenance of clones of Ag-selected memory B cells81– if becoming a PC is the default pathway following memory B-cell activation and Ag clearance, it is unclear how memory B cells are retained to seed subsequent responses. Furthermore, if severing the interaction with T cells is the molecular trigger for differentiation, Ag-specific Ig will only be made once Ag has been eliminated.

We have attempted to resolve these discrepancies by proposing a stochastic division-based model39,64 that allows for (1) the early production of Ig without the requirement of a trigger for differentiation, such as the withdrawal of CD40L, and (2) the maintenance and expansion of the pool of memory B cells (Fig. 2). When memory B cells are activated with T-cell help, especially IL-10, they begin dividing and a proportion differentiate into ISCs at each division, with this proportion increasing with each subsequent division. Two populations of ISC can be discriminated by the differential expression of CD38, with CD38 cells representing precursors of CD38+ plasmablasts (Fig. 2 and refs39,70). The signals required for sustaining division and survival change once memory B cells differentiate into ISCs. Removing CD40L leads to arrest of proliferation and apoptosis of the CD38 B-cell lineage, which contains ISCs and undifferentiated B-blasts. In contrast, IL-2 and IL-10 sustain proliferation and survival of the CD38+ ISCs while CD40L becomes dispensable, thus allowing for a continued increase in the number of CD38+ ISCs (Fig. 2). Cells that remain undivided, or have only undergone limited division, and are therefore undifferentiated, may correspond to ‘memory-stem’ cells81 capable of responding to subsequent Ag encounter. An attractive explanation for the stochastic features of both Ig isotype switching and ISC differentiation described above is that expression of transcription factors alters the probability of differentiation occurring.82 Indeed, we have observed increased expression of B lymphocyte induced maturation protein-1 (Blimp-1) and X-box binding protein-1 (XBP-1), which are highly expressed in PCs,82 and decreased expression of B cell lineage-specific activator protein (BSAP)/Pax5, which is expressed at all stages of B-cell development except PCs,82 in mouse and human B cells that have differentiated in vitro into cells with an ISC phenotype (ref.31 and unpublished data). Our model explains the previous data of Arpin and colleagues,71 which supported a role for removal of CD40L as a trigger for differentiation, in the following way: by removing CD40L, CD38 B-cell blasts become highly susceptible to cell death, while the survival of CD38+ cells is less compromised. Thus, when the proportion of CD38+ CD20± and CD38 CD20+ cells are measured in culture, the former will predominate in the absence of CD40L, leading to the interpretation that expansion of ‘plasma’-like cells requires the withdrawal of T-cell help.

Figure 2.

A stochastic division-linked model of plasma cell generation. Memory B cells stimulated with T cells begin dividing and continue to divide as long as T-cell-derived signals, including CD40L, are maintained. Memory B-cell blast expansion therefore requires frequent serial contact with T cells. Throughout the life of the memory blast, a stochastic mechanism is operating that directs the transition to an immunoglobulin-secreting cell (ISC) (either CD38 or CD38+) according to a regulable probability. In early divisions, the probability is low; however, it increases with consecutive divisions, resulting in a greatly enhanced rate of plasmablast formation later in the response (shown by thickening arrows). CD38+ ISCs (i.e. plasmablasts) are no longer dependent on CD40L, and therefore do not require serial contact with T cells. Consequently, these cells can proliferate in the presence of T-cell-derived cytokines only (e.g. IL-10). Presumably these plasmablasts have altered migration properties and home to sites where further plasma cell (PC) maturation occurs. Plasmablasts and their early expansion contribute to the initial, rapid antibody response. The long-lived, Ag-independent non-dividing PCs are responsible for persistent antibody well after the Ag is removed.

Mouse B cells

In contrast to human B cells, murine naïve cells can readily be stimulated to become ISCs in culture using TD stimuli such as IL-4 and CD40L.31 However, similar to isotype switching by naïve human and mouse B cells, and the generation of ISCs from human memory B cells, the frequency of secreting cells increases with progressive divisions.31 Thus, as for human memory B cells,39 a stochastic division-linked process was suggested where the probability of becoming a secreting cell can be altered by a number of cytokines.31

In the mouse, the progressive increase in both switching and ISC generation with division allowed these processes to be monitored simultaneously.31 The frequency of cells that underwent both differentiation events was consistent with independent stochastic processes operating within each cell. This experimental result strongly supports our stochastic view of differentiation and provides a relatively accessible new paradigm for exploring the generation of heterogeneity within populations of stimulated cells.31

Regulators of differentiation

Our models of B-cell activation and differentiation allowed us to re-examine the effects of many known regulators and assign them to one or other of the composite steps involved in generating ISCs from naïve and memory B cells.


  • 1IL-4 and IFN-γ. IL-4 enhances the rate of proliferation of naïve mouse B cells, while simultaneously promoting isotype switching to IgG1 and IgE.26,27,32 IFN-γ produces a weaker, but still positive, enhancement of proliferation while inducing efficient switching of cells in late divisions to IgG2a.32 When the two cytokines were added to cultures together, an interesting mode of antagonism was observed between them: IFN-γ reduced the ability of IL-4 to promote B-cell proliferation, resulting in a proliferation behaviour similar to that observed with IFN-γ alone.32 IFN-γ also inhibited division-linked switching to IgE and appeared to effectively replace this isotype with IgG2a in a dose-dependent manner. Curiously, IFN-γ had little effect on IL-4-induced switching to IgG1.32 Thus, these studies illustrate how separate events can be independently regulated by external signals, and how potentially simple rules can be established to define how they will work in combination.
  • 2TGF-β. TGF-β profoundly inhibits proliferation at the same concentrations it promotes class switching to either IgG2a or IgA by murine B cells stimulated with LPS or CD40L, respectively.28 Consequently, only a small number of switched cells are seen in cultures with this cytokine. This raises the question of how the percentage of IgA+ cells in sites such as Peyer's patches can reach as high as 50%.21 Given the relation to division, we proposed that sustained exposure to gut-derived Ag would drive some cells, possibly those of highest affinity, to divide sufficient times despite the strong antiproliferative effect of TGF-β.28 Thus, the presence of this cytokine at sites of continual Ag exposure would serve both as a suppressor of clonal expansion and as a facilitator of switching to the secretory Ig isotype.
  • 3IL-5. IL-5 promotes a large increase in the amount of Ig secreted in cultures of murine B cells.19,83 In a comprehensive analysis of its effects, it was found that IL-5 solely affected the probability of becoming an ISC associated with each division.31 This illustrated that internal division-linked processes can be regulated with remarkable specificity.
  • 4IL-6 and IFN-α. IL-6 has long been known to enhance survival and Ig secretion by normal and malignant human PCs84 as well as plasmablasts generated in vivo75 or in vitro.78,85 IFN-α has also been reported to enhance Ig production by human B cells86,87 and is important for the maintenance of long-term humoral immunity.88 In a recent study it was proposed that a specialized population of dendritic cells (DCs), plasmacytoid DCs (pDCs),89 contributes to the differentiation of human B cells into ISCs by the sequential production of these two cytokines.90 First, IFN-α, produced in large quantities by stimulated pDCs, induces the production of plasmablast precursors. Secondly, IL-6 acts on these precursor cells to yield high-rate Ig-secreting PCs.90 It is likely that these cytokines achieve this end-point by altering the rate of proliferation, and subsequently differentiation, of activated memory B cells.
  • 5B-cell activating factor belonging to the tumour necrosis factor (TNF) family (BAFF). BAFF has been found to play a critical role in B-cell homeostasis by acting as a survival factor for developing B cells.91 BAFF exerts its effect by binding to one of three different receptors: BAFF-R, B cell maturation antigen (BCMA), or transmembrane activator and calcium modulator and cytophilin interactor (TACI).91 Exposure of activated memory B cells to BAFF was found to increase the number of ISCs generated, especially those secreting IgA.40 BAFF had no effect on the proliferation or phenotype of CD38+ effector cells, but it did reduce their rate of apoptosis, demonstrating that BAFF achieved its effect by acting as a survival factor.40 Interestingly, expression of BCMA was induced on CD38+ B cells, while BAFF-R was down-regulated from these cells, suggesting that BCMA may be the receptor that delivers survival signals to plasmablasts.40 BAFF is produced by DCs,91,92 and the amount produced can be increased by CD40L and IL-10.92 Thus, interactions between CD40L-expressing/IL-10-producing follicular helper T cells93 and activated B cells within lymphoid tissues in response to specific Ag would generate CD38+ plasmablasts (Fig. 3). Concomitant interactions between such T cells and DCs would induce production of BAFF, which may act with IL-10 to facilitate survival of Ag-specific plasmablasts (i.e. CD38+ ISCs; Fig. 3). This model may also explain the previous findings that DCs can potently modulate the differentiation of human B cells into ISCs.77,92,94
Figure 3.

B-cell activating factor belonging to the tumour necrosis factor (TNF) family (BAFF) supports the survival of CD38+ plasmablasts: a schematic outline of how interactions between T cells, memory B cells and dendritic cells (DCs) contribute to facilitate the generation of immunoglobulin-secreting cells (ISCs). The same T-cell-derived signals that drive B-cell proliferation and differentiation (i.e. CD40L, IL-2 and IL-10) stimulate DCs to produce BAFF (and also IL-10). Thus, DC-derived BAFF may act as a survival factor for de novo generated ISCs, which lose dependence on CD40L, while undifferentiated memory B-blasts remain dependent on T-cell help for their survival. BCMA, B cell maturation antigen.

BcR cross-linking

Many Ags first engage and cross-link the Ag receptor on the B-cell surface. The resulting signal induced by Ag can be mimicked by anti-Ig reagents.9,19 When included in culture, different anti-Ig reagents varied in their ability to promote B-cell proliferation.95 However, all were capable of inhibiting divison-linked Ig isotype switching to IgG1.95 This result suggests that physical properties of the Ag, such as the degree of multivalency, can be detected and help regulate the division-linked control of differentiation events, such as isotype switching.


Coculture of activated human B cells with transfectants expressing the CD27 ligand CD70 yielded an increased frequency of CD38+ CD20+ cells, leading to the suggestion that interactions between CD27+ memory B cells and activated CD4+ T cells induced to express CD70 may favour the differentiation of human B cells into PCs.96,97 It is currently unknown whether CD27–CD70 interactions increase the frequency of CD38+ CD20+ cells by enhancing proliferation of these differentiated cells, or of their precursors, or by improving their survival. Ligation of CD27 on human B cells by recombinant CD70 has very little effect on cell proliferation,98 suggesting that it may act as a survival factor for ISCs, akin to BAFF. We propose a scenario whereby initial T-cell–B-cell interactions mediated by CD40L and CD40 induce proliferation and differentiation of B cells into ISCs, a proportion of which become CD40L-independent, and subsequent interactions between CD27 and CD70 may maintain this population of plasmablasts by improving their survival. This is consistent with the temporal expression of CD40L and CD70 on activated T cells, with CD40L being maximally expressed after 24 hr of stimulation, and decreasing thereafter, and CD70 expression increasing after 2 days and reaching peak levels after 7 days.97

Striking a balance: optimizing clonal expansion and development of antibody

Our division-based model of immune regulation depicted in Figs 1 and 2 has a number of important theoretical implications. For example, during the secondary immune response a balance must be struck as the formation of ISCs occurs at the expense of the memory B-cell pool.81 The division-linked, stochastic model of ISC development supports the view that memory B cells operate as renewable ‘stem cells’ which are not depleted by continued exposure to Ag.81 By regulating the proportion of ISCs formed per division, a balance between increasing the number of memory ‘stem’ cells and rapidly generating ISCs, and therefore antibody, can be achieved. In Fig. 4, a model of differentiation is presented to illustrate how division-linked changes in differentiation rate can help optimize the rapid development of antibody while preserving a memory capability. Figure 4(a) shows the effect of having the proportion of undifferentiated cells that become ISCs in each division follow a normal distribution, as determined experimentally for both human39 and mouse31 B cells, and the resulting number of memory ‘stem’ cells versus ISCs at various times is plotted in Fig. 4(b–e). The maximum frequency in Fig. 4(b–d) is altered progressively from 10 to 60% at division 6. The sequence shows that, at a peak rate of ISC generation of 10%, memory blast expansion is rapid, and greatly exceeds the number of ISCs being formed (Fig. 4b). At 20% (Fig. 4c), and more noticeably at 40% maxima (Fig. 4d), the memory population still expands significantly, but remains capable of generating a greater number of ISCs earlier in the response. At 60% there is a rapid increase in ISC numbers; however, this result is achieved at the expense of expanding the memory compartment (Fig. 4e). The above model can be contrasted with one that assumes a constant rate of differentiation per division (Fig. 4f–j). When the two models are compared, it is clear that a short period of memory blast expansion before differentiation enables greater numbers of ISCs to develop within a shorter time than can be achieved with a constant rate of differentiation. While this example necessarily simplifies the immune response, it is reasonable to conclude that regular exposure to infectious diseases during evolution has optimized the rate of division-linked differentiation to deal efficiently with pathogens in the short term through rapid development of ISCs, while ensuring preservation and enhancement of the memory pool for protection against future exposure.

Figure 4.

Optimizing the rate of immunoglobulin-secreting cells (ISC) formation while preserving the memory compartment. The stochastic, division-linked model of ISC formation is presented under different values for the maximum level and contrasted with a linear model of differentiation. (a) The frequency of ISCs formed by the end of each division number is given as a normal distribution with a mean division number of 6 and a standard deviation of 1·5 divisions. The maximum frequency at division 6 is varied and set to either (b) 10%, (c) 20%, (d) 40% or (e) 60%. The numbers of memory blasts (○) and ISCs (•) are calculated before each consecutive division and plotted as shown (b–e) from a starting number of 104 cells. The ISCs are not set to divide in this calculation to better illustrate their time of appearance. Division time in this model is synchronized and set arbitrarily at 40 hr for the first division, and at 15 hr for each subsequent division. (f) The frequencies of ISC formation for the ‘linear’ model of differentiation, where the rate per division is held constant. The results of the linear model with the frequency of memory blasts forming ISCs per division set at (g) 10%, (h) 20%, (i) 40% or (j) 60% are shown.

This model also serves to illustrate a counter-intuitive conclusion concerning vaccines. During immunization with an artificial, non-replicating Ag, the production of antibody itself is not essential, and may diminish the strength of the response by clearing Ag and reducing the phase of T-cell-dependent memory B-cell growth. By contrast, a strategy that diminishes the rate of PC differentiation may enhance the expansion of the memory pool, because fewer cells become ISCs, and the introduced Ag will take longer to be cleared. Therefore, locally regulating ISC differentiation during immunization may boost memory B-cell numbers while reducing the rate of antibody production, thereby enhancing long-term immunity to natural infections. It is highly likely that such manipulation is possible because different vaccines have been found to have disparate effects on the levels of circulating antibodies as well as Ag-specific memory B cells. Specifically, subsequent booster immunizations with tetanus toxoid dramatically increased the levels of circulating antitetanus antibodies, yet had only a marginal effect on the frequency of tetanus-specific memory B cells, while similar booster immunizations with diphtheria toxoid caused a large increase in the frequency of Ag-specific B cells, but had no effect on serum levels of diphtheria-specific antibody.99

Somatic hypermutation

Somatic hypermutation (SHM) is a process utilized by B cells to produce a high-affinity repertoire of Ag-specific B cells elicited in response to TD Ag. SHM generally occurs within GCs located in follicles of secondary lymphoid tissues, and precedes the induction of Ig class switch recombination.100,101 It has been proposed that the extensive proliferation of GC B cells is associated with somatic mutation.102 By analysing somatic mutation in GC-derived B cells (i.e. GC B cells themselves, memory B cells or PCs), we observed that the frequency of mutations within the Ig V region genes expressed by isotype-switched B cells generally exceeded that of IgM-expressing cells (Table 1; compiled from data published in refs48,60,61,100,101,103–110 and unpublished data). Although the mutational load in PCs producing IgM and IgA from human intestine was greater than that in the corresponding PCs found in the spleen, the increased frequency of mutation in class-switched Ig V region genes still exceeded that of non-switched cells (Table 1). On the basis of our in vitro data describing the relationship between cell division and Ig isotype switching, it can be predicted that IgG+ and IgA+ B cells will have undergone more rounds of cell division than IgM-expressing memory B cells or PCs. Thus, as mutations accumulate with division, this observation is consistent with the proposal that switched cells will generally have divided more times than unswitched cells.

Table 1.  Somatic mutation increases with Ig isotype switching
B cellsIg isotypeMutation frequency (mean percentage ± SD)
  1. The frequency of mutation of the Ig V region expressed by naïve B cells, GC B cells, memory B cells or PCs was calculated based on refs48,60,61,100,101,103–110 (and unpublished data). The data for naïve B cells were pooled from papers reporting values for cells isolated from peripheral blood, spleen, tonsils and bone marrow; the data for GC B cells are from tonsils only.

NaïveIgM+ IgD++0·17 ± 0·28
Germinal centreIgM+1·93 ± 1·61
IgG+3·78 ± 2·5
Blood memoryIgM++ IgD+3·81 ± 1·0
IgG+/A+5·2 ± 0·8
Spleen memoryIgM++ IgD+2·54 ± 1·83
IgG+/A+5·71 ± 1·0
Splenic plasma cellsIgM+2·5 ± 1·7
Intestinal plasma cellsIgM+6·31


From the above discussion, a picture emerges in which interleaving internal stochastic processes can be modulated further by external signals to yield tremendous variations in the B-cell response. While there is much work still to do, especially in understanding the molecular control of the interleaving stochastic events and the ways in which division itself might help to promote or diminish subsequent differentiation events, we believe that the discovery of the importance of division in this process will help to elucidate a rich new framework for exploring otherwise extremely complex processes.


This work was supported by the National Health and Medical Research Council (NHMRC) of Australia. SGT is the recipient of an R. D. Wright Biomedical Career Development Award from the NHMRC; PDH is a Senior Research Fellow of the NHMRC.