In the course of an adaptive immune response, the humoral arm of the immune system generates distinct types of antigen-specific B lymphocytes in secondary lymphoid organs: memory B cells form a persisting population of antigen-specific B cells that can become activated and quickly differentiate into plasma cells upon re-exposure to the same antigen . Terminally differentiated antibody secreting plasma cells constitute effector cells of humoral immunity by functioning as immunoglobulin (Ig) secreting cells . They can become long-lived and persist in specialized survival niches in the bone marrow for extended periods of time (several years in humans), thereby secreting antibodies constantly. Together, these different B cell populations ensure an initial defense as well as long-lasting protection against antigens upon rechallenge, thus marking an important part of immunologic memory against pathogens. However, long-lived plasma cells can also be harmful by secreting autoreactive antibodies.
Memory B cells and plasma cells of high affinity are derived from highly organized histological structures in secondary lymphoid tissue called germinal centers (GCs) [3-6]. Mature GCs can be divided into different zones: the dark zone (DZ) at the base of the GC (proximal to the T cell zone), and the light zone (LZ) at the apical pole of the GC, characterized by the presence of follicular dendritic cells (FDCs) . The border between the GC area and the naïve B cells in the follicular mantle is termed outer zone . The GC constitutes a site where several crucial decisions on B cell fate are made: After extensive proliferation and somatic hypermutation , the B cells undergo a selection process based on the affinity of their BCR . This is thought to involve competition for antigen (Ag) captured on the surface of follicular dendritic cells (FDCs) . Only B cell receptors harboring high affinities for a given antigen survive, whereas lower affinity B cells undergo apoptosis, ultimately resulting in affinity maturation of the humoral immune response . After the selection process, a second decision step follows for a B cell: It can either stay in the GC to enter a further round of division , or it leaves the GC. Exiting B cells can either become memory B cells, or they terminally differentiate into antibody secreting plasma cells .
Over the past few years, a model of the complex transcriptional regulatory network for terminal differentiation of B cells into antibody secreting cells has been developed. The earliest known molecular step involves down regulation of the transcription factor Pax5 [14, 15]. The transcriptional repressor Pax5 is required for maintenance of the identity of mature B cells , and is also expressed in memory B cells. Pax5 negatively regulates Ig production by repressing Igh  and Igk  transcription, expression of J chain  and XBP-1 . Loss of Pax5–by an unknown regulator- is sufficient to induce the onset of IgM secretion [14, 21]. On the other hand, Pax-5 promotes expression of BACH2 , which support the formation and functions of GCs such as class switching [23, 24].
Bcl-6 is another transcriptional repressor, which is highly expressed in GC B cells  as well as in T follicular helper cells . It is required for the formation of GCs [27-29] and promotes GC functions such as proliferation as well as class-switch recombination and somatic hypermutation. Additionally, Bcl-6 inhibits plasma cell differentiation in GCs by suppressing Blimp-1  in association with MTA3 [31, 32].
Another crucial transcription factor controlling not only GC formation but also differentiation into plasma cells is IRF4 [33, 34]. IRF4 seems to act in a dose-dependent way: It promotes GC formation and functions, e.g., class switch recombination when abundant in low concentrations. At higher concentrations it can trigger plasma cell differentiation. IRF4 expression is controlled and kept at low levels in B cells by MITF, hence MITF-deficient B cells spontaneously differentiate into antibody secreting cells . In cooperation with STAT3, IRF4 up-regulates the expression of Prdm1, the gene encoding Blimp-1 in an IL-21-dependent mechanism .
Blimp-1, the master regulator of plasma cell differentiation, silences genes expressed in mature B cells, including Pax-5 , class II transactivator and CXCR5. Blimp-1 is also able to repress Bcl-6 [14, 21, 38]. On the other hand, Blimp-1 up-regulates the expression of XBP-1, which induces genes involved in the regulation of protein synthesis and secretory pathways that are important for the secretion of large amounts of antibodies . Importantly, Blimp-1 is not necessary for the earliest steps in plasma cell differentiation, including the onset of immunoglobulin secretion . Maturation from the plasmablast stage to long-lived bone marrow ASCs is accompanied by an increase in Blimp-1 expression [40, 41] and it is actually required for the maintenance of long-lived bone marrow plasma cells. Notably, Blimp-1 is not expressed in memory B cells [42, 43].
Locationand Migrationof Antibody Secreting Cells Within SecondaryLymphoid Organs
Inhibition of Bcl-6 by Blimp-1 [and vice versa ] explains why only very few Blimp-1+ plasma cells can be detected inside GCs of mice and humans by histology [42, 44, 45]. However, CD138+ antibody secreting cells have been detected to be located adjacent to GCs, lining the border between GC dark zone and T cell zone . It is this region where, in the course of an immune response, the first cognate interactions between B and T cells happen . After activation, B cells upregulate CCR7 (the receptor for T cell zone chemokines CCL19/21) and are directed to the T-B border . In this region, the initial interactions of B cells with T cells take place, subsequently leading to the formation of GCs. Intravital imaging has revealed dark zone B cells in mature GCs to move in a directed fashion towards the T-B border . The presence of plasmablasts at this location [Fig. 1 and ] further underlines the importance of this region, not only during the initiation but also during the maintenance of the GC response.
The majority of antibody secreting cells within secondary lymphoid organs are located in lymph node medullary cords (Fig. 1), and in extrafollicular foci of the splenic red pulp. In histology, accumulations of plasma cells can be observed reaching from the follicles throughout the T cell zone to the medullary cords (Fig. 1).
Two-photon studies have revealed GC B cells to be highly motile . In contrast, at least a fraction of plasma cells in medullary cords are thought to be rather sessile [50, 51]. Upregulation of the plasma cell-specific transcription factor Blimp-1, which goes along with terminal differentiation, i.e., transition from the plasmablast to the plasma cell stage  negatively correlates with their motility . In line with this, the term “plasmablast” is now widely used for a migratory antibody secreting cell, although it historically refers to an antibody secreting cell that still has the capacity to divide.
How is the translocation from plasmablasts generated in the GCs to medullary cords and extracellular foci mediated? Plasmablasts down-regulate CXCR5  and thus lose their capacity to migrate towards high concentrations of the B cell chemokine CXCL13 in vitro . Instead, they migrate towards CXCL12, a chemokine that is highly expressed in medullary cords and the splenic red pulp . However, CXCL12 is also abundant in the dark zone of GCs [56, 57] and it is not completely clear how the migration from the GCs to the CXCL12-rich environment in extrafollicular sites of secondary lymphoid organs is regulated. Expression of EBI2 is up-regulated on plasma cells  and this 7-transmembrane receptor has been shown to direct B cells to the outer follicle areas [58, 59] and extrafollicular foci . However, Fooksman et al. have shown that pertussis toxin was not able to block the migration of plasmablasts from B cell follicles to lymph node medullary cords in vivo. This indicates that G protein-coupled receptor signaling is not involved in this translocation, thereby ruling out a critical role for EBI2 as well as for chemokine receptors and the receptors for sphingosine-1-phosphate. When compared to naive B cells and GC B cells, plasma cells display a more directed pattern of migration [Fig. 1 and Ref. ]. Fooksman et al. suggested this type of migration pattern provides an efficient way to translocate from follicles to medullary cords even without chemotactic signals. Alternatively, the influence of other, yet undefined chemotactic cues could account for an active migration to the medullary cords.
Extrafollicular plasma cells are thought to be mainly short-lived as a high degree of apoptosis has been shown to occur among them ; however, there are also reports about the persistence of long-lived plasma cells in the spleen . Notably, long-lived plasma cells have been shown to occur in secondary lymphoid organs of NZW/B mice, a model for systemic lupus erythematosus . Moreover, chronically inflamed organs have also been shown to harbor long-lived antibody secreting cells [64-66]. However, the physiological site for the maintenance of long-lived plasma cells is the bone marrow.
Plasmablast Migrationfrom Secondary Lymphoid Organsto the Bone Marrow
A fraction of plasmablasts generated in systemic immune responses leave the secondary lymphoid organs and travel via the blood to the bone marrow. These plasmablasts are generally thought to originate from GC responses. However, recent data challenge this view: Long-lived plasma cells have been shown to be induced in a T-independent response as well as after inhibiting the formation of GCs by blocking CD40L in vivo . The exact routes, which these cells take to exit from spleen and lymph nodes have not yet been defined, but it is known that egress of antibody secreting cells from the spleen and from Peyer's Patches requires signaling through S1P1 [68, 69]. S1P, the ligand for S1P1 is produced by erythrocytes and therefore highly abundant in the blood stream , plasmablasts in the blood express higher amounts of S1P1 as compared to their counterparts in secondary lymphoid organs. Treatment of mice with the immunosuppressant S1P antagonist FTY720 inhibits plasmablast egress into the blood in vivo . In addition, beta-2 Integrin has been shown to play a role in plasmablast exit from peripheral lymph nodes .
Plasmablast entry pathways into the bone marrow are not yet fully elucidated. Presumably plasma blasts enter via sinusoidal veins in the bone marrow, but the mechanisms that mediate their transmigration have not been investigated in detail. One possible candidate involved into this process may be CD62L, as plasma blasts have been reported to express this homing receptor . Another candidate is the adhesion molecule CD22, which is expressed on plasmablasts and becomes downregulated in terminally differentiated plasma cells. Sialylated ligands for CD22 have been reported to be present on sinusoidal endothelial cells of the bone marrow .
Clearly, efficient homing of plasmablasts to the bone marrow depends on CXCR4, the receptor for CXCL12, a chemokine which is produced in high amounts by bone marrow stromal cells, hence its name stromal derived factor 1 (SDF-1) . Plasmablasts lacking CXCR4 fail to reach the bone marrow and instead accumulate in the blood, suggesting a crucial role for this receptor on these cells in entering the bone marrow from the blood . Interestingly, the capacity of splenic plasmablasts to migrate to CXCL12 ex vivo is restricted to a tight time window, during the first week of a systemic immune response , since they have been shown to lose their migratory capacity between day 8 and 12 after boost immunization. In line with this, the number of antigen-specific plasmablasts is found to peak in the blood of immunized humans at day 7 after boost . How other chemokines act on bone marrow plasma cells to fine-tune their localization and retention in the parenchyma remains to be investigated. Potential candidates are ligands of CCR2, as antigen-specific antibody titers were significantly reduced in mice with plasma cells deficient for this chemokine receptor . A fraction of plasmablasts has also been shown to migrate to ligands of CXCR3 , suggesting a role of this chemokine receptor in the accumulation of antibody secreting cells in inflamed organs. In fact, chronically inflamed kidneys of NZB/W lupus mice have been shown to contain CXCR3+ antibody secreting cells  and a role for this chemokine receptor in the entry of plasmablasts into the inflamed CNS has been demonstrated in a mouse model of viral encephalomyelitis .
Dynamicsof Mucosal Plasma Cells
The largest fraction of antibody secreting cells in the body is actually located in mucosal associated lymphoid tissue (MALT). Virtually all plasma cells induced in mucosal immune reactions secrete antibodies of the IgA isotype. Class switch of B cells to IgA is induced in isolated lymphoid follicles, Peyer's Patches and mesenteric lymph nodes. These tissues provide an environment rich in factors like transforming growth factor beta (TGFߚβ) and a proliferation inducing ligand (APRIL), which have been shown to promote IgA responses . A polymeric immunoglobulin receptor-dependent mechanism allows dimeric IgA to be transported via the gut epithelium into the lumen, where IgA is present mainly as a part of the mucus layer covering the surface . Mucosal IgA+ plasmablasts first use lymph vessels to migrate to the mesenteric lymph nodes, and then they enter the blood circulation through the thoracic duct and home back to the lamina propria. The chemokine receptors CCR9 and CCR10 on plasma blasts are responsible for homing to the small and large intestine, respectively. Retinoic acid plays a role in imprinting of the gut homing mechanism of IgA+ plasmablasts by inducing CCR9 . The adhesion molecule a4ß7 is also mediating gut tropism in plasmablasts . Besides that, a fraction of IgA+ plasma cells in the diffuse lamina propria has been proposed to be generated locally, without the need for recirculation . The majority of plasma cells in the blood of humans in steady state are derived from mucosal immune responses . By FACS, up to 50% of bone marrow plasma cells in mice have been shown to be IgA+, but these values have to be interpreted with caution since the bone marrow contains large blood-filled sinusoids, therefore one cannot rule out a contribution of blood-derived plasma cells. While IgA+ plasma cells have been shown to express CXCR4 and migrate to the bone marrow , it is not known to what extent they can contribute to the resident long lived plasma cell compartment. Antibody secreting cells in the lamina propria have been thought to be short-lived  but there is indirect evidence in the recent literature that there might be a long-lived compartment present in this tissue as well [85, 86]. A recent study has described that some lamina propria plasma cells are able to produce tumor necrosis factor alpha (TNFα) and inducible nitric oxide synthase (INOS), revealing an unexpected heterogeneity in the antibody secreting cell compartment in the gut .
Bone Marrow Survival Niches: A Specialized Stromal Microenvironment Fostering Long-Lived Plasma Cells
The abundance of plasma cells in the bone marrow was already recognized in the 1970s . Twenty years later the group of Radbruch and coworkers first showed that the murine bone marrow contains a fraction of long-lived plasma cells . Slifka et al. made a similar observation using irradiation of mice  and showed that the lifespan of plasma cells can reach values that are similar to the lifetime of a mouse. The independence of the long-lived plasma cell compartment from replenishment through memory B cells was later confirmed using different strategies of memory B cell depletion [90, 91].
The bone marrow is thought to sustain the survival of plasma cells in specialized microanatomical niches. Plasma cells have been shown to depend on a combination of extrinsic survival factors to survive  and their colocalization with several cell types, which produce these survival factors has been demonstrated by histology: plasma cells in the bone marrow are in close contact with reticular stromal cells (Fig. 2). These stromal cells have been shown to produce the chemokine CXCL12 . CXCL12 seems to act in two different ways on plasmablasts versus plasma cells: as a chemokine, which has been shown to regulate plasma immigration into the bone marrow [54, 55], but also as a potent plasma cell survival factor as it has been demonstrated in vitro . Besides its function in plasma cell migration and survival, it also plays a crucial B cell development and in the formation of the hematopoietic stem cell niches and lymphoid progenitor niches [94, 95]. Whether there is a heterogeneity among the CXCL12-producing stromal cells, i.e., whether the same stromal cell type can form a plasma cell niche or a hematopoietic stem cell niche has not yet been addressed. The isolation of bone marrow stroma cells for flow cytometric characterization is a technical challenge and markers like VCAM-1 or laminin, which are used to label bone marrow stromal cells in histology are also part of the extracellular matrix, making the interpretation of the stainings difficult. The recently reported generation of fluorescent reporter mice for different stromal cell populations in the bone marrow will help to overcome the challenge of visualizing stromal cells in vivo .
Accessory Cellsof the Plasma Cell Survival Niches
In addition to stromal cells, a number of cell types have been suggested in the literature to play an additional role as accessory cells for plasma cell maintenance. Colocalization of bone marrow plasma cells with megakaryocytes producing interleukin (IL)-6 APRIL has been demonstrated by histology . IL-6 has been shown to support plasma cell survival in vivo  and in vitro . The TNF-receptor family member APRIL binds to BCMA and TACI on plasma cells and mediates survival on plasma cells [98, 99]. Its function on plasma cells is mainly mediated via signaling through BCMA , which recently has been shown to induce high expression of mcl-1, a member of the bcl-2 family of pro-survival molecules . Induction of mcl-1 crucially depends on BCMA-signaling in plasma cells from the bone marrow but not from the spleen, underlining the role of APRIL in the survival of long lived plasma cells.
Mice deficient for the megakaryocyte growth factor receptor Cmpl have 8-fold reduced numbers of megakaryocytes and plasma cells are reduced to one-third. Another study found 50–70% of bone marrow plasma cells in the vicinity of APRIL+ IL-6+ eosinophils . Mice lacking eosinophils displayed a 70% reduction of long-lived bone marrow plasma cells in vivo, suggesting that the eosinophils had provided survival signals for these plasma cells. Activation of eosinophils has been shown to trigger the expression of plasma cell survival factors and their number in the bone marrow was increased after secondary immunization, suggestive of a previously unrecognized role of this cell type in immunological memory . Of note, the half-life of eosinophils is only a few days under physiological conditions, but it has been shown to be prolonged in the case of infections . It is currently not clear whether this is also the case for eosinophils in the plasma cell niches, allowing for a prolonged supply of APRIL for the plasma cells.
Another group demonstrated that myeloid precursor cells are actually the main producers of APRIL in the bone marrow . Basophils have also been shown to support plasma cell survival in vitro and in vivo, partially through the production of IL-4 and IL-6 .
Ly-6Chi monocytes , and DCs  have also been proposed to play a role in plasma cell survival in mouse models. Bone marrow DCs have been shown to directly contact plasma cells, and they are able to support plasma cell survival via CD80 and CD86, which interact with CD28 on plasma cells . Notably, stimulation with anti-CD28 antibodies in vitro resulted in activation of the prosurvival NF-kB-pathway and prolonged survival in plasma cells from the bone marrow, while it had no effect on splenic plasma cells. Whether this effect can be attributed to different stages of maturation or to the existence of distinct subsets of antibody secreting cells is not clear. Crosslinking of CD80/86 by CD28 on plasma cells in addition induced the production of IL-6 in bone marrow DCs, this in turn enhanced immunoglobulin production in the plasma cells. Notably, plasma cells in the lymph node have also been shown to instruct myeloid cells to produce APRIL and IL-6 .
Finally, osteoclasts have been shown to support human plasma cell survival in vitro . Except for stroma cells, all of the aforementioned cell types are of hematopoietic origin, therefore one would expect a high degree of turnover, leading to a replacement of these cell types in the niches. This raises the question how plasma cells are continuously provided with survival factors from the transient cells over extended periods of time. There are several facts, which suggest that long lived plasma cells in the bone marrow are sessile, although this has not yet been demonstrated directly. They express high levels of Blimp-1  and in antibody secreting cells in the lymph nodes Blimp-1 expression levels have been shown to negatively correlate with cellular motility . Furthermore, plasma cells lose their capacity to migrate to chemotactic signals after their arrival in the bone marrow . Together, these findings suggest that the plasma cell niche consists of static parts, which organize the niche and control its composition (e.g., reticular stroma cells) as well as dynamic components (e.g., hematopoietic cells, as most of the accessory cell types that have been described to be a part of the niche have a lifespan that is much shorter than that of a long-lived plasma cell). A stable stromal cell population of defined size, which provides a limited amount of niche space by supplying survival factors and attracting other, transient accessory cells to the niche may function to control the size and homeostasis of the long-lived plasma cell pool. In humans, antigen-specific HLA-DRhi plasmablasts that are on their way to the bone marrow can be detected one week after immunization . However, in parallel also the frequency of mature (HLA-DRlo) plasma cells in the blood increases, leading to the hypothesis that sessile long-lived bone marrow residents can be outcompeted and replaced in their niches by incoming, motile blasts . Assuming a random exchange of plasma cells in the bone marrow by influxing newly generated plasmablasts, computational modeling has implicated that it would take more than 500 years in humans to reduce the number of antigen-specific bone marrow plasma cells generated in one immune response if an individual undergoes on average four infections per year . However, this model does not explain that antigens differ in their capacity to elicit longevity of plasma cells in the bone marrow. Whether this reflects variations between different immunizations in the number of generated plasma cells that reach the bone marrow or whether the plasma cells show an intrinsic difference in their capacity to survive is not clear.
Direct evidence for such a competition model is not yet available. In the future, it will therefore be crucial to investigate the impact of dynamics of the survival niches (i.e., cell motility and turnover of all cellular niche components). Intravital multiphoton imaging in the bone marrow has already been successfully applied to understand the biology of perivascular immune niches  as well as stem cell niches , and it will help to understand the dynamics of plasma cell niches.
Morphologic data describing the localization of plasma cells in the bone marrow have been derived from static histology. These data have been very helpful in understanding the structure of the bone marrow; however, colocalization analyses, which do not take the variation of frequencies and cell size distributions of different bone marrow cell populations into account, cannot discriminate between random or nonrandom colocalization events. Systems biology approaches, which simulate the distribution of the different bone marrow cell populations, may help to solve this question in the future. First of all, the crucial cell-interactions of the niche components have to be identified. It might be that plasma cells can induce or enhance the secretion of survival factors in the niche cells.
One also has to take into account that many of the plasma cell survival factors are soluble, so a direct contact between the cell providing the factor and the consuming plasma cell might not even be required to deliver the survival signals. In the human lamina propria, heparan sulfate proteoglycans on stromal cells have been shown to bind APRIL produced by neutrophils and therefore create a local microenvironment for plasma cells where this survival factor is enriched . It has been shown for multiple myeloma cells that CD138, a surface molecule, which is also expressed by nonmalignant plasma cells can act as a coreceptor for APRIL , providing a mechanism how these cells catch soluble survival factors. A more detailed analysis of the motility of the niche cells as well as the distribution of the soluble survival factors will be needed to solve this question.
On the other hand, adhesion molecules like VLA-4 and LFA-1 are essential for plasma cell survival in the bone marrow. Administering a combination of blocking antibodies against these two molecules in vivo resulted in a 75% reduction of bone marrow plasma cells . Ligands for VLA-4 that are present in the bone marrow include VCAM-1, fibronectin, and osteopontin, LFA-1 binds to ICAM-1,-2,-3. Both VLA-4 and LFA-1 probably act by fixing the plasma cells into their niches. Another molecule that has been suggested to promote survival of plasma cells by functioning as an adhesion molecule is CD93 .
Plasma Cellsas Therapeutic Targets
Autoantibodies are known to play a crucial role in the pathogenesis of several autoimmune diseases. Consequently, the depletion of long-lived autoreactive plasma cells is a promising therapeutic option, especially under conditions where targeting of B cells via CD20 has proven to be inefficient, e.g., in refractory systemic lupus erythematosus (SLE) or rheumatoid arthritis (RA) . One obvious option is to interfere with the signals promoting plasma cell survival, and the development of Atacicept (TACI-Ig), a fusion protein of the human immunoglobulin constant region and TACI, the soluble receptor for the B cell survival factor BAFF and the plasma cell survival factor APRIL, has shown promising results in two independent phase 2 clinical trials for RA [118, 119]. In the future it will be a challenge to translate the increasing knowledge about the bone marrow plasma cell survival niche into new therapeutic strategies to interfere with these niches in order to deplete autoreactive plasma cells.
Another strategy for plasma cell depletion is targeting them intrinsically. The proteasome inhibitor Bortezomib has been shown to efficiently deplete short- and long-lived autoreactive plasma cells in two different mouse models for SLE. Bortezomib acts via activating the terminal unfolded protein response (UPR) . Plasma cells produce enormous amounts of immunoglobulins [up to 10,000 molecules/s and cell ], which inevitably results in the accumulation of un- or misfolded proteins, and which renders them especially susceptible for apoptosis after the regular function of the proteasome is inhibited. Importantly, Bortezomib treatment was not only able to prevent the onset of the disease but it also ameliorated disease after the onset of clinical symptoms by reducing autoantibody titers, a finding which is of relevance especially for a future translation into the clinics. However, there are side effects that have to be considered when using Bortezomib, such as peripheral neuropathy , therefore the development of proteasome inhibitors, which more specifically target antibody secreting cells is desirable for the future. The recent finding that plasma cells crucially depend on autophagy for their homeostasis and longevity may also open alternative options for targeting these cells in the future . In summary, all the currently available therapeutic options have in common that they do not distinguish between pathogenic, autoreactive plasma cells, and “normal” protective plasma cells, conferring protective antibodies, which are equally depleted. The long-term goal therefore should be the development of therapies, which specifically target autoantibody-producing plasma cells in the individual patient as a step to a more personalized medicine.
In summary, a lot of progress has been made in unraveling the complex transcriptional network that governs the differentiation of B cells into plasma cells. Intravital microscopy has helped to understand the migration and interactions of B cells and plasmablasts in secondary lymphoid tissues. Long-lived plasma cells are located in the bone marrow but we are only beginning to understand their lifestyle. In the future it will be necessary to further analyze them in the tissue context in order to fully understand the mechanisms, which keep them alive in their survival niches.
This project is funded by the Deutsche Forschungsgemeinschaft (DFG HA5354/4ߚ1 to A.E.H.) and JIMIߚa DFG core facility network grant for intravital microscopy. The authors thank Sabine Gruczek, Patrick Thiemann, and Manuela Ohde for assistance with animal care and Robert Günther for excellent technical assistance. We thank the lab of J.J. Lee (Mayo Clinic, Scottsdale, AZ) for providing the anti-MBP antibody. We are grateful to A. Miyawaki, RIKEN Brain Institute, Japan for the Cdt1-mKO cell cycle reporter mice and to S. Nutt, WEHI Melbourne, Australia for Blimp-1:GFP reporter mice.