This study provides evidence of an essential role for the DZ in the GC reaction. A major difference between the LZ and DZ is the presence in the LZ of immune-complex-decorated FDC and the majority of GC Tfh cells. The cues that promote the transition of centroblasts to centrocytes had not been examined in detail before now, but the increased expression by centrocytes of genes commonly associated with acute activation was thought to suggest exposure of centroblasts to centrocyte-inducing signals upon arrival in the LZ (Victora and Nussenzweig, 2012
). However, our data are not consistent with such a model because we found nearly normal proportions of CXCR4-deficient GC B cells displaying a CD83lo
centroblast-like phenotype, despite their sequestration in the LZ. Furthermore, access to the DZ was not essential for efficient passage through the S to G2-M checkpoint of the cell cycle. Together, these findings strongly suggest that centroblast and centrocyte functions are limited to a certain cellular stage, rather than to a particular site. We propose a model where switching from the centroblast proliferative stage to the centrocyte selection stage progresses according to a cellular “timer” that operates independently of DZ-derived signals. Activation of the timer might be triggered by LZ-derived signals such as receipt of T cell help. According to this model, decreases in surface CXCR4 expression by GC B cells are coordinated with reductions in proproliferative and SHM gene expression and increases in CD83 and CD86. This phenotype transition causes the cell to migrate into the LZ, rather than occurring as a consequence of it. Expression of centrocyte-associated genes (such as CD86) might change the nature of the interaction with T cells, thereby limiting the capacity for positive selection to the period after the current round of mutation and clonal expansion is complete.
While the decision of when to exit the DZ centroblast stage might be set by a cellular timer, regulation of reentry into the centroblast compartment probably involves cells testing their new BCRs by competing for the formation of productive T cell interactions (Allen et al., 2007b
). Delivery of peptide antigen to a subset of GC B cells was sufficient to stimulate increased cellular division, centroblast differentiation, and ongoing GC participation at the expense of their neighbors (Victora et al., 2010
). Two recent studies indicate that c-Myc might play an important role in DZ cyclic reentry because it is required for continued participation in the GC, and it is expressed by a small fraction of centrocytes that are enriched for high affinity Ig rearrangements and that have recently entered the synthesis phase of cell cycle (Calado et al., 2012; Dominguez-Sola et al., 2012
). By supporting a model in which T cell-derived signals induce transient c-Myc expression, leading to reentry to the DZ stage, these studies seem consistent with transient c-Myc expression being involved in resetting the centroblast “timer.” While we emphasize the role of a zone-independent and thus likely intrinsic cellular program in coordinating the centroblast gene expression profile and in regulating the transition back to the centrocyte stage, extrinsic factors such as the nature of earlier T cell interaction might imprint aspects of behavior while in the DZ-associated state.
Although DZ access is not essential for centroblast differentiation and proliferation, it is required for effective competition and continued participation within the GC. We propose two non-mutually exclusive models that might explain this. First, the polarized GC might reflect a need for the temporal and spatial separation of centrocyte and centroblast functions, rather than cues in each zone promoting them. The most compelling example for this might be the physical separation of SHM and selection; this might facilitate complete exchange of existing BCR with newly encoded protein from the mutated locus prior to antigen and T cell exposure. Premature entry into the LZ might drive aberrant BCR signaling or lead to inappropriate T cell interactions that could result in negative selection (clonal deletion) or premature positive selection. For example, failure to upregulate CD86 following BCR crosslinking in anergic B cells leads to their killing by FasL-expressing T cells (Rathmell et al., 1998
), and the low CD86 expression by centroblasts might cause these Fashi
cells to suffer a similar fate following antigen presentation in the GC LZ. Over time, improper GC B cell selection would manifest as a decrease in somatic mutation accruement. As a second possibility, the activity of the SHM machinery might be bolstered by cues only present in the DZ. We did not observe differences in AID mRNA or protein abundance in CXCR4-deficient cells; however, these measures do not exclude the possibility that SHM activity is lower; its regulation is particularly stringent and includes transcriptional, posttranscriptional, and posttranslational mechanisms (McBride et al., 2004
). Less efficient acquisition of nucleotide substitutions would be expected to lead to less frequent improvements in affinity and reduced competitiveness in the GC.
A key function of the GC is to generate memory B cells and PCs that provide protection against future infections of the same or similar kinds. We found that the frequency of CXCR4-deficient PCs closely matched that of the concurrent GC population, suggesting that differentiation toward this fate was not negatively affected by their inability to respond to CXCL12. This finding was true regardless of whether we looked early in the response, when many antibody-secreting cells will be generated via an extrafollicular response, and after 4 weeks of infection when most PCs should be generated in the GC. However, it must be noted that we could not assess the efficiency of generating long versus short-lived PCs because CXCR4 is required intrinsically within PCs for BM homing (Hargreaves et al., 2001; Nie et al., 2004
). In contrast to PC generation, the representation of CXCR4-deficient cells within the memory compartment was greater than expected given their GC participation defect, consistent with the idea that a transition to the memory cell fate might be favored when cells cannot access the DZ. It is important to note that currently available tools do not allow for the definitive identification of memory cells that have recently exited the GC reaction, especially during complex polyclonal responses; however, work has suggested that costaining with CD73 and BrdU should enrich for such a developmental stage (Anderson et al., 2007; Kaji et al., 2012; Taylor et al., 2012
). The signals that cause more cells to adopt this fate are not clear, but we think it might again be a consequence of CXCR4-deficient GC B cells engaging with T cells prematurely or aberrantly due to their positioning in the antigen and T cell rich LZ. Previous studies have indicated that the selection checkpoint for memory cell generation might be less stringent than it is for PCs (Smith et al., 2000; Victora et al., 2010
). An increase in memory cell generation has been reported in at least two other settings where the B-T interaction is abnormal; in mice lacking interleukin-21R (IL-21R) specifically on B cells (Zotos et al., 2010
) and in mice lacking Fas on all class-switched B cells (Hao et al., 2008
). Increased memory cell numbers were also seen in Bcl6
heterozygous mice, which might be analogous to where T cell-derived IL-21 cannot drive the maintenance of Bcl6 protein levels (Kaji et al., 2012
) (Linterman et al., 2010
). Therefore, we speculate that a possible increase in memory cell generation by CXCR4-deficient B cells might reflect their receiving a quality of T cell help that is sufficient to rescue them from deletion but that cannot drive their PC differentiation or stimulate their continued participation in the GC.
GC B cells are acutely dependent upon trophic factors present in their microenvironment; cells displaced from that setting die within a matter of hours (Wang et al., 2011
). While LZ FDCs provide one source of such factors, our confocal analysis of CXCL12-GFP and of Ubi-GFP mice revealed the DZ to contain a patchwork of tight and highly branched reticular cells that might also contribute. Previous studies had noted some VCAM-1, fibrinogen, and CD35 staining in this zone but had not examined the nature of the stromal network in detail (Allen and Cyster, 2008
). We observed some variation in DZ coverage by CXCL12-producing stroma; in some GCs, the network extended throughout the CD35lo
region but formed patches of “labyrinth-like” structures, seemingly similar morphologically to the reticular network described in human basal LZs (Imal and Yamakawa, 1996
). LN GCs mostly form with the LZ proximal to the subcapsular sinus from where antigen drains and the DZ expands to border the neighboring T or medullary compartment. This raised the possibility that CXCL12-expressing reticular cells (CRCs) in the DZ arise by the GC engulfing the neighboring compartment as the follicle expands. However, we also identified a T-zone proximal network of CXCL12-expressing reticular cells in primary follicles. Like DZ CRCs, primary follicle CRCs were mostly distinct from neighboring FRCs both in their morphology and in their not strongly associating with ERTR7 staining, although it should be noted that both primary and secondary follicles also contained ERTR7+
perivascular cells. We therefore think it more likely that preexisting follicular CRCs help to establish early stages of GC polarization and later form DZ CRCs as the GC matures. An important question for future study is the relative contributions of CRCs, CXCL12-expressing perivascular cells, and GC-proximal FRCs to GC organization and maintenance. In this regard, it is interesting to note that CXCL12-expressing perivascular and endothelial cells are important in organizing and maintaining hematopoietic stem cell and committed B cell progenitor niches in the BM (Tokoyoda et al., 2004
). In this setting, cells compete for cues derived from the stromal network. It will be interesting to see whether a similar fitness competition plays out in the GC.
In summary, we have provided evidence that centrocyte differentiation from centroblasts occurs as part of a timed cellular program. We believe a key function of this program is to temporally separate the processes of SHM and mitosis from selection; as the molecular program transitions the cell to the centrocyte stage, proliferation and mutation genes are reduced and the cell readies itself for optimal engagement with antigen and with T cells. Therefore, much of GC behavior that has traditionally been described as reflecting a function of the LZ or the DZ instead occurs as part of the particular stage of the GC B cell program. However, we also provide strong evidence that the spatial separation of LZ and DZ functions is critical for maintenance of effective GC responses.