Plasma cell S1P₁ expression determines secondary lymphoid organ retention versus bone marrow tropism

To test the impact of FTY720 treatment on ASC homing during the primary response, mice were immunized with the T cell–dependent antigen, NP-CGG in the adjuvant alum, and then at day 4 and 7 with FTY720. The frequency of ASCs in spleen and BM was then determined at day 7 and 10 (Fig. 1 A). The FTY720 treatment was delayed until day 4 to limit effects on the initial recruitment of recirculating antigen-specific B and T cells. The IgG ASC frequency in the spleen was little affected by the FTY720 treatment (Fig. 1 A). In contrast, there was a decrease in the number of IgG ASCs in the BM at day 7 and 10 (Fig. 1 B). Similarly, s.c. immunization led to appearance of IgG ASCs in the draining LNs, and FTY720 treatment did not greatly affect this response. However, FTY720 largely inhibited the appearance of IgG ASCs in the BM after s.c. immunization (Fig. 1, A and B).

Efforts to determine whether there was an effect on ASC egress from lymphoid organs into the blood were hampered by the low numbers of ASCs entering into the blood during the primary response. We therefore performed secondary immunizations, treating the mice with FTY720 at day 1.5 after secondary immunization since BM homing via the blood occurs by day 3 of the secondary response (7). Again the treatment had little effect on the IgG ASC number in the spleen (Fig. 1 A). A baseline of NP-specific IgG ASCs were detected in the BM because of the primary immunization, but the number was increased twofold by day 3 and four- to fivefold by day 7 in the control group. FTY720 treatment inhibited the increase in BM ASC number (Fig. 1 B). IgG ASCs were readily detectable in the blood of control mice at day 3 and were largely gone by day 7 (Fig. 1 C). In contrast, few if any IgG ASCs appeared in the blood of FTY720-treated mice (Fig. 1 C). We also asked whether FTY720 could affect BM homing after ASCs had entered the blood, by adoptively transferring spleen cells from mice at day 3 of the secondary response into unimmunized recipients that were either untreated or treated 1 d before with FTY720. ELISPOT analysis 1 d after transfer revealed similar numbers of donor-derived NP-specific ASCs in the BM of control and FTY720-treated mice (Fig. 1 D). Thus, FTY720 treatment inhibits the migration of IgG ASCs from spleen into blood but does not inhibit the homing of adoptively transferred IgG ASCs from blood to BM.

S1P₁ is abundant in naive lymphocytes and is required in these cells for egress from lymphoid organs (17, 18). S1P₁ is also abundant in endothelial cells (22). The ability of FTY720 to inhibit cell egress from lymphoid organs may be caused by agonistic effects on endothelial cells, down-modulation of receptors on immune cells, or disruption of S1P gradients (23–25). To test whether there was an intrinsic requirement for S1P₁ in ASCs for egress we reconstituted lethally irradiated Igh^b mice with S1P₁^−/− Igh^a fetal liver cells and small numbers of Igh^b WT BM cells as a source of WT T cells. After a 6-wk reconstitution, the mice were primed and then boosted with NP-CGG in alum. In mixed chimeric animals that mounted similar day 3 splenic IgG ASC responses (Fig. 2 A), very few S1P₁^−/− ASCs could be detected in the BM at day 3 or day 7 (Fig. 2 B). Consistent with S1P₁ being required for ASC egress from the spleen, very few IgG ASCs could be detected in the blood at day 3, whereas IgG ASCs could be detected in the blood of WT chimeras (Fig. 2 C).

Immunohistochemical analysis revealed that the distribution of IgG ASCs in FTY720-treated mice at day 5 of the secondary response was similar to their distribution in saline-treated controls (Fig. 3 A). Side-by-side comparison of the distribution of S1P₁^−/− IgG1^a and S1P₁^+/+ IgG1^b ASCs in the spleen of mixed chimeras further demonstrated that this pathway is not required for positioning of IgG ASCs in clusters within the RP (Fig. 3 B).

As a read-out of S1P₁ function we examined the ability of ASCs to migrate in response to S1P. We compared the response of IgG ASCs from spleen at day 3 versus day 5 of the secondary response because previous studies have established that the majority of ASCs egress at day 3 with few cells continuing to egress into blood at day 5 (1). Day 3 but not day 5 splenic IgG ASCs showed a small but significant (P < 0.05) chemotactic response to S1P (Fig. 4 A). ASCs at both time points demonstrated a similar responsiveness to SDF-1 (Fig. 4 A). Consistent with previous demonstrations that high dose FTY720 treatment ablates S1P responsiveness in lymphocytes, cells from the treated mice had lost the ability to respond to S1P while continuing to respond to SDF-1 (Fig. 4 A). We also tested the ability of IgG ASCs from the blood to chemotax to S1P. Two technical difficulties with this experiment were the very small numbers of ASCs present in blood and the presence of large amounts of S1P in blood that down-regulate and desensitize S1P₁ (26). In preliminary experiments, we found it difficult to wash the cells free of blood S1P. However, when the CD45.1 blood cells were mixed with a 10-fold excess of CD45.2 spleen cells at the time of isolation and washing and then the spleen cells depleted using antibodies to CD45.2, it was possible to reveal an S1P response in the blood cells, and this response exceeded that of spleen ASCs isolated in parallel and through the same steps (Fig. 4 B).

An assessment of the migratory response to S1P was also performed using cells from immunized Blimp1^GFP reporter mice (27). Blimp1 is up-regulated at an early step in ASC differentiation and is essential for plasma cell development and maintenance (28). As previously observed, the ASCs induced in the spleen at day 3 of the secondary response were predominantly Syndecan1⁺ and GFP^int (Fig. 5 A), whereas there was an accumulation of Syndecan1⁺ GFP^hi cells at later time points (27). The ASCs in the blood at day 3 were consistently Syndecan1⁺ GFP^int, and very few Syndecan1⁺ GFP^hi cells were detected (reference 27 and unpublished data). In migration assays, the splenic GFP^int cells showed a migratory response to S1P, whereas the GFP^hi cells showed little response (Fig. 5, A and B). The splenic GFP^int cells also showed a robust response to SDF-1 (Fig. 5, A and B). These findings confirm that the day 3 splenic ASC population contains S1P-responsive cells.

Analysis of ASCs that accumulate in E- and P-selectin–deficient mice showed that B220^int/lo cells were more chemokine responsive than B220^neg cells (29). Our observations agree with these findings because the migratory GFP^int cells are B220^int/lo, whereas the poorly migratory GFP^hi cells express little if any B220 (Fig. 5 A, bottom and reference 27). In addition, within the GFP^int population, we saw a trend toward more S1P responsiveness in ASCs with higher amounts of B220 (Fig. 5 A, bottom). However, B220 levels were not markedly different between GFP^int cells in spleen and blood (unpublished data). In the course of these experiments we observed that CXCR4 was expressed at intermediate levels on the surface of blood ASCs compared with splenic ASCs (Fig. 5 C). Despite the lower CXCR4 expression, the magnitude of the blood and splenic ASC SDF-1 chemotactic responses appeared similar (Fig. 4 B). A lack of direct correlation between ASC CXCR4 levels and SDF-1 chemotactic responses has been observed previously (29–31).

To measure S1P₁ abundance in ASCs in blood versus spleen, Blimp1^GFP reporter mice were primed and boosted with NP-CGG and Syndecan1⁺ GFP^int and GFP^hi spleen cells or Syndecan1⁺ GFP⁺ blood cells were purified at day 3 of the response. The GFP^int subset in the spleen was further divided into CXCR4^hi and CXCR4^int to explore the possibility that the CXCR4^int population contained cells that were the most closely related to the blood ASCs (Fig. 5 C). Compared with naive B cells, GFP^hi splenic ASCs had ∼20-fold reduced S1P₁ transcript levels (Fig. 6 A). Splenic GFP^intCXCR4^hi cells also showed a marked reduction in S1P₁ expression, whereas GFP^intCXCR4^int cells showed a trend toward increased expression (Fig. 6 A). GFP^int ASCs sorted from the blood had amounts of S1P₁ that were within twofold of the levels detected in splenic B cells (Fig. 6 A). Expression of S1P₂, S1P₃, S1P₄, and S1P₅ in blood ASCs was low or undetectable (unpublished data). We also quantitated Blimp1 transcripts and found that they were undetectable in splenic and blood B cells and abundant in the ASCs as expected (unpublished data). Finally, we examined the abundance of KLF2 (also known as lung KLF), a transcription factor that can promote S1P₁ transcription (32, 33), and found that levels were reduced in splenic GFP^hi and GFP^int CXCR4^hi ASCs (Fig. 6 B). As for S1P₁, KLF2 transcripts were more abundant in the splenic GFP^intCXCR4^int subset than in GFP^intCXCR4^hi cells and levels in blood GFP^int ASCs were similar to splenic B cells (Fig. 6 B).

The above findings demonstrate that egress of IgG ASCs from secondary lymphoid organs into blood and BM is sensitive to inhibition by FTY720 treatment and strongly dependent on intrinsic S1P₁ expression. S1P₁ abundance and S1P responsiveness is found to be low in most splenic ASCs but is detectable within a Blimp1^int CXCR4^int splenic ASC population and is most enriched in ASCs isolated from the blood. These findings establish an important role for S1P₁ in IgG plasma cell homing and they suggest that differential regulation of S1P₁ expression in differentiating plasma cells may be a key factor determining whether they remain in secondary lymphoid organs or exit these organs to home to other sites, such as the BM. We also find differential expression of KLF2 between splenic and blood ASCs, pointing to a possible role for this transcription factor in defining the gene expression program associated with the homing decision of differentiating plasma cells.

In a previous study, treatment with FTY720 from the day of primary immunization was found to reduce the magnitude of the IgG plasma cell and germinal center response (21). This may have been a consequence of reduced recruitment of antigen-specific T cells from other sites caused by the generalized block in secondary lymphoid organ egress (34). We delayed treatment of the mice for 4 d in the primary response or 1.5 d in the secondary response to provide an opportunity for recruitment of circulating antigen-specific B and T cells, and we found that the secondary lymphoid organ ASC responses were of similar magnitude in the treated and untreated groups, indicating that treatment did not severely affect the ability of the animals to mount an antibody response. In agreement with this conclusion, production of neutralizing antibodies after the first 3 wk of lymphocytic choriomeningitis virus and vesicular stomatitis virus infection were unaffected by low dose (0.3 mg/kg) FTY720 treatment (20). The ability of S1P₁-deficient B cells to mount antibody responses also demonstrates that B cell expression of this receptor is not essential for B cell activation and differentiation. However, because of limitations in the mixed fetal liver/BM chimera approach, our experiments have not examined the efficiency of each stage of the antibody and memory B cell response. Future experiments with mice carrying floxed alleles of S1P₁ (17) and B cell–selective Cre recombinase will be needed to more fully address the impact of ASCs and B cell S1P₁ deficiency on maintenance of long-term antibody production and on B cell memory.

The mechanism by which FTY720 reduces lymphocyte egress from lymphoid organs is under active investigation (18, 35), and it has been debated whether FTY720 impacts lymphocyte egress from the spleen (14). Our findings here provide evidence that egress of IgG ASCs from the spleen can be inhibited by FTY720 treatment. Although the pathway by which cells exit the spleen is not defined, it is likely that the cells need to reach RP sinusoids to return to circulation. Whether the ASCs that normally egress from the spleen correspond to cells that have first traveled into the RP is not yet clear. However, like spleen and LN ASCs, blood ASCs are found to have down-modulated CXCR5 and CCR7 (and up-regulated CXCR4), making it likely that these cells migrate out of the white-pulp before exiting into circulation (6). The lack of CXCR5 and CCR7 expression on these cells but the continued requirement for S1P₁ to permit efficient egress indicates that the intrinsic function of S1P₁ cannot solely be to overcome the retention function of these chemokine receptors. However, it remains possible that CXCR4 normally participates in retaining cells in the spleen and LNs, and one function of S1P₁ may be to overcome CXCR4-mediated retention. In this regard it is notable that the Blimp1^GFP+ cells in the blood were CXCR4^int compared with the bulk of the Blimp1^GFP+ cells in the spleen. We propose that the reduced CXCR4 expression together with the increased S1P₁ expression in a subset of cells contributes to shifting the balance of signals in favor of S1P₁ and promoting egress. However, CXCR4 is only partially down-modulated, consistent with it also being required for lodgment of ASCs in the BM. Beyond this possible interplay, we anticipate that S1P₁ will be required during egress to overcome additional retention signals, provide directional information, and/or promote reverse transmigration across blood or lymphatic vessels. Although we have focused on the mobilization of IgG ASCs to the BM, we anticipate that similar requirements will exist for IgA ASC egress from mucosal lymphoid tissues before their homing to epithelial surfaces.

Our findings suggest that definition of the mechanism by which a subset of differentiating ASC up-regulate S1P₁ may illuminate how ASCs choose between the secondary lymphoid organ versus BM (or mucosa) tropic plasma cell fate. This cell fate decision may also encompass becoming short versus long lived, though some studies have indicated that long-lived plasma cells (PCs) are present in secondary lymphoid organs and BM (36) and, reciprocally, some of the PC homing to BM may be short lived. Although our studies show that S1P₁ is necessary, our studies have not established that this receptor is sufficient to determine whether an ASC undergoes egress or is retained, and additional differences between the cells may be important. The finding that KLF2, a transcription factor shown to directly induce S1P₁ expression in T cells (32, 33), is increased in blood Blimp1^GFP+ cells is consistent with a role for this factor in promoting S1P₁ expression in ASCs. KLF2 regulates the expression of additional genes in T cells (32, 33) and it may regulate expression of further genes involved in the commitment of differentiating B cells to becoming BM tropic ASCs. Previous studies have suggested that BM tropic ASCs are induced predominantly from germinal center B cells or memory B cells (5, 37–39). Indeed, the synchronized wave of ASCs traveling through the blood at day 3 of the secondary response most likely reflects rapid differentiation of memory B cells into BM tropic ASCs. Although it is thought likely that these ASCs contribute to the pool of long-lived PCs, we have not established that this is the case for the cells tracked in the present study. Finally, although we were able to enrich based on intermediate CXCR4 and Blimp1^GFP expression for splenic ASCs with an egress-related phenotype, this cell population had lower amounts of S1P₁ and KLF2 than the cells isolated from blood, suggesting that we had only achieved partial enrichment. The future development of reagents that permit tracking of S1P₁ expression at the level of single B cells and ASCs should provide a method to isolate and further characterize the earliest cells that are taking on the BM tropic ASC fate.

C57BL/6 (B6) and B6 CD45.1 mice were from the National Cancer Institute. Blimp1^GFP/+ mice were generated as described (27). S1P₁^+/− mice (16) were crossed for at least five generations to B6 or 129 mice, and then S1P₁^+/− mice were intercrossed. S1P₁^+/+ or S1P₁^−/− fetal liver cells were prepared at embryonic day 12.5 and genotyped by PCR (16). Each preparation was mixed with B6 BM cells (1.5 × 10⁵ and 3 × 10⁶ cells, respectively) and transferred to reconstitute three lethally irradiated B6 CD45.1 mice (6). Mice were immunized 6 wk after reconstitution. For FTY720 treatment, B6 mice were injected i.p. with 3.0 mg/kg of FTY720 or with an equivalent volume of saline 4 and 7 d after primary and 1.5 d after secondary immunization. FTY720 was from a custom synthesis performed at SRI International. Animals were housed in a specific pathogen-free facility and all experiments were in accordance with protocols approved by the University of California San Francisco Institutional Animal Care and Use Committee.

B6 mice were immunized with 50 μg of alum-precipitated NP-CGG (Solid State Sciences) i.p. or s.c. in a volume of 200 μl. In some experiments, mice received a secondary immunization of the same type 3 wk later. Spleen, LNs, blood, and BM were harvested from mice at the indicated time points as described previously.

Flow cytometric data were collected on a FACS LSRII (BD Biosciences) and analyzed using FlowJo software (Tree Star). Antibodies used were: biotin-conjugated anti-CXCR4 (2B11; BD Biosciences), PE-conjugated anti-Syndecan (281–2; BD Biosciences), APC-conjugated anti-B220 (RA3-6B2; BD Biosciences) and NP (40), and APC-Cy7–conjugated anti-CD4 and CD8 (GK1.5, 53–6.7; BD Biosciences). Streptavidin-PE–Cy5.5 was from Caltag, and DAPI was used to exclude dead cells. Spleens of immunized B6 mice treated with or without FTY720 were frozen in Tissue-Tek OCT compound (Baxter Scientific). Cryostat sections (7 μm) were fixed in acetone and stained as described previously (41) with sheep anti-IgD (Binding Site, Inc.) and biotin-conjugated rat anti-IgG1, IgG1a, and IgG1b (MOPC-31C, 10.9, B68-2; BD Biosciences). Sheep antibodies were detected using HRP-conjugated donkey anti–sheep IgG (Jackson ImmunoResearch Laboratories). Biotinylated reagents were detected using the SA-AP (Jackson ImmunoResearch Laboratories). Enzyme reactions were developed with conventional substrates for peroxidases (diaminobenzidine; Sigma Chemical Co.) and alkaline phosphatase (Fast Blue; Sigma Chemical Co.).

Spleen cells from B6 mice and Blimp1^GFP mice were stained, and ASCs and B cells were isolated with a MoFlo cell sorter (Cytomation). To isolate blood Blimp1^GFP+ cells, blood from 15–20 mice at day 3 after secondary immunization was used for each experiment. In some cases, to generate enough mice bearing the Blimp1^GFP allele in their hematopoietic cells, BM chimeras were made by transferring CD45.2 Blimp1^GFP BM into irradiated CD45.1 B6 recipients and allowing 6 wk for reconstitution before primary immunization. The purity of the sorted cells was typically >95%. RNA was prepared from sorted cells with RNeasy (QIAGEN), and cDNA was used for quantitative PCR on an ABI 7300 sequence detection instrument (Applied Biosystems) using the following primer and probe sets with AmpliTaq Gold DNA polymerase (Applied Biosystems) CXCR4, forward, CGGCTGTAGAGCGAGTGTTG and reverse, TCTCCAGACCCCACTTCTTCA; probe FAM-CATGGAACCGATCAGTGTGAGTATATACACTTCTGA-TAMRA; Blimp1(Prdm1): forward, GGATCTTCTCTTGGAAAAACGTGT and reverse, AGCCGTGTAAAGTAGACTGCCTTG; probe FAM-TACGACCTTGCCAAAGGCTGCATTTTAAA-TAMRA; and HPRT, S1P₁, S1P₂, S1P₃, S1P₄, and S1P₅ as reported (18). KLF2 (and HPRT) was quantitated using SYBR Green (Bio-Rad Laboratories), and KLF2, forward, TATCTTGCCGTCCTTTGCCA and reverse, TTAGGTCCTCATCCGTGCC primers.

Cells were tested for transmigration across uncoated 5-μm transwell filters (Corning Costar Corp.) for 3 h to S1P (Sigma-Aldrich), SDF-1 (PeproTech), or medium in the bottom chamber and were enumerated by ELISPOT or, in the case of cells from Blimp1^GFP/+ mice, by flow cytometry (42). To prepare blood cells for chemotaxis, 2 × 10⁶ Ly5.1 Igh^a RBC-lysed blood cells isolated by cardiac puncture were mixed with 10⁸ Ly5.2 Igh^b RBC-lysed spleen cells, incubated for 30 min at 37°C in DMEM 0.5% fatty acid–free BSA, washed twice, and then the Ly5.1 and Ly5.2 cells were separated using Ly5.2 and AutoMACS. Because of the low frequency of ASCs, the blood cells that migrated across 8 transwells and the spleen cells that migrated across 4 transwells in response to S1P were pooled and tested in the ELISPOT or FACS assays.

Freshly isolated spleen, blood, BM, and LN cells were used in ELISPOT assays as described previously (6). Plates were precoated with NP₂₅-BSA (Solid State Sciences). IgM, IgM^a, IgM^b, IgG1^a, IgG1^b, and IgG were detected with anti–mouse IgM biotin (Caltag), IgM^a biotin (DS-1; BD Biosciences), IgM^b biotin (AF6-78; BD Biosciences), IgG1 biotin (BD Biosciences), IgG1^a biotin (BD Biosciences), and IgG1^b biotin (BD Biosciences), followed by SA-HRP and True Blue (Kirkegaard & Perry Labs, Inc).

We thank Chris Allen, Shuwei Jiang, and Susan Watson for help with FACS sorting and Tri Phan for comments on the article.

K. Kabashima was a HHMI postdoctoral associate and is currently funded by the Takeda Science Foundation. N.M. Haynes was supported by a CJ Martin Fellowship. J.G. Cyster is a HHMI Investigator. This work was supported in part by grants from the National Institutes of Health (AI40098 and AI45073).

The authors have no conflicting financial interests.