KLF2 expression in IgG plasma cells at their induction site regulates the migration program

Plasma cells can be found in a variety of tissues across the body that can be divided into two sites—where they differentiate from activated B cells (induction site) and the tissues where they migrate to robustly secrete their antibodies (effector tissues) (Tarlinton et al., 2023). For instance, after the generation of short-lived plasma cells (SLPCs) in secondary lymphoid organs, some of these cells enter niches in the bone marrow, wherein they mature and survive as long-lived plasma cells (LLPCs) (Manz et al., 1997; Slifka et al., 1995; Kallies et al., 2004). Since LLPCs are a key cellular compartment for the durability of antibodies (Halliley et al., 2015), they are important for protection from reinfection, thereby underpinning successful vaccine development.

LLPCs are thought to reside in specialized survival niches, such as in the bone marrow, proposed originally by MacLennan and colleagues (Sze et al., 2000). This term currently describes the collection of cells and the factors they produce that provide survival signals to plasma cells. Based upon such a concept, the characterization of plasma cells and identification of bone marrow niche cells have been intensively examined. Detailed analyses of plasma cells have revealed heterogeneity of gene expression and metabolism that may be associated with heterogeneity in survival (Duan et al., 2023; Liu et al., 2022; Shi et al., 2015; Kallies et al., 2004). As for bone marrow niches, the collection of candidate cells colocalized with LLPCs, and the factors that provide survival signals to LLPCs are becoming clearer (Robinson et al., 2020).

Despite such progress, the question of what the key determinant(s) for plasma cell longevity are has remained unanswered. In this regard, two (not mutually exclusive) conceptual models have been put forward (Tarlinton et al., 2023). First, newly generated SLPCs simply migrate and lodge in appropriate niche-containing effector tissues by serendipity. Assuming the existence of heterogeneity of niche composition, SLPCs accessing niches highly capable of supporting plasma cell survival can mature, thereby acquiring superior longevity (effector tissue model). The existence of such heterogeneity of niche cells has been recently suggested (Lindquist et al., 2019; Chang et al., 2018). An alternative “induction site model” postulates that heterogeneity of SLPCs may be created at their generation site, for instance, depending on the amount of T cell help, innate signals, and/or the microenvironment. Such variables may then instruct the longevity of plasma cells. Supporting the induction site model, differences in the form of the vaccine antigen and/or adjuvants are likely to primarily affect the plasma cell generation processes, providing them with distinct longevity (Bhattacharya, 2022). Indeed, a repetitive antigen is known to induce more durable protective antibodies than a non-repetitive one (Slifka and Amanna, 2019). However, neither model has yet been directly tested.

Here, to address the mechanism by which some plasma cells become long-lived, we first characterized and compared the newly generated plasma cells in secondary lymphoid tissues and those that just arrived in the bone marrow. We found that among the two populations of newly generated plasma cells that we identified in secondary lymphoid tissues, integrin β7^hi and β7^lo cells, only β7^hi cells were prone to egress and home to the bone marrow. The bone marrow–tropic plasma cells expressed higher level of the KLF2 transcription factor, which we show is required for promoting their egress by upregulating S1PR1 and CD11b. Disruption of plasma cell egress by KLF2 ablation resulted in defects in antibody durability and protection against influenza reinfection. Hence, the KLF2 expression level in IgG plasma cells at the induction site is one of the key factors that determine antibody durability by regulating their migration program.

To compare the characteristics of newly generated plasma cells in the spleen and those subsequently arriving in bone marrow, we employed germinal center (GC) fate-mapping mice (S1pr2-creERT2/tdTomato) (Shinnakasu et al., 2016). After adoptively transferring 4-hydroxy-3-nitrophenyl acetyl (NP)-specific B1–8^germS1pr2-creERT2/tdTomato B cells into congenically marked mice, we immunized them with NP-CGG in alum i.p., treated them with tamoxifen (Fig. 1 A), and analyzed donor-derived tdTomato⁺CD138^hiTACI^hi plasma cells in the spleen or bone marrow (Fig. S1 A). Because they were analyzed 4 days after tamoxifen treatment, the tdTomato⁺ plasma cells are recently generated and GC experienced. We were particularly interested in examining by flow cytometry (FCM) the status of cell surface molecules involved in trafficking, such as adhesion receptors, G-protein–coupled receptors, and related cell surface molecules (Fig. 1 B). Before the flow experiments, shown in Fig. 1 B, we performed pre-screening to identify candidate molecules that differ in their overall expression between splenic and bone marrow plasma cells (Fig. S1, B and C).

Among various adhesion and G-protein–coupled receptor molecules, the expression pattern of integrin β7 and CD11b (integrin αM) was unique (Fig. 1 B); in contrast to the bimodal expression pattern (particularly of integrin β7) seen on splenic plasma cells, β7^hi and CD11b^hi plasma cells were predominantly present in the bone marrow. In addition, overall expression levels of CD11c (integrin αX), CXCR4, and CD62L were somewhat lower on bone marrow plasma cells.

To determine how early these two plasma cell populations (integrin β7^hi and β7^lo) are generated after transition from GC B cells, we did time-course experiments after tamoxifen injection (Fig. S2, A and B), demonstrating that both populations are generated at very early time points after GC labeling. This phenomenon was also manifested in GC-independent plasma cells (Fig. S2, C and D). Moreover, these two populations were not only observed in B cell receptor (BCR) transgenic settings, since we obtained essentially similar results in non-BCR transgenic mice (Fig. S2, E and F). Importantly, although the spleen can be considered as both an induction site and an effector tissue (Tarlinton et al., 2023), consistent with previous homeostatic and NP-specific polyclonal studies (Koike et al., 2023; Robinson et al., 2022), the half-life of B1–8^germ plasma cells in bone marrow was longer than that in spleen (Fig. S3, A and B).

The simple interpretation of the above data is that T-dependent immune responses generate plasma cells, consisting of integrin β7^lo and β7^hi populations in the spleen, the latter of which is prone to migrate to the bone marrow. This interpretation was reinforced by the following transcriptomic analysis. By employing the Fig. 1 A adoptive transfer experiments, we analyzed gene expression by using barcode-based digital RNA sequencing (RNA-seq) (Jin et al., 2023) As shown in Fig. 1 C, principal component analysis (PCA) demonstrated that the splenic integrin β7^lo and β7^hi plasma cell fractions are transcriptionally distinct and that the bone marrow plasma cell fraction is more akin to the β7^hi splenic one.

To further strengthen the above interpretation, we examined clonal relatedness between integrin β7^hi/β7^lo splenic and β7^hi bone marrow plasma cell fractions by using tdTomato⁺NP⁺IgG1⁺ cells at day 10.5 after NP-CGG immunization of non-BCR transgenic S1pr2-creERT2/tdTomato mice (Fig. 1 D). After cloning the antibody sequences from single-cell sorted tdTomato⁺NP⁺ plasma cells, we analyzed V_H186.2 sequences and assessed their CDR3 sequences. Then, “clonal distance” was determined, showing that the bone marrow plasma cell fraction contains more sequences close to the β7^hi splenic plasma cell fraction (Fig. 1 E).

We further conducted i.v. plasma cell transfer experiments, demonstrating that integrin β7^hi donor plasma cells maintained integrin β7 expression, whereas their integrin β7^lo counterparts did not acquire integrin β7 expression in the bone marrow (Fig. S3, C and D). Taken together, these findings strongly support the conclusion that integrin β7^hi bone marrow plasma cells originate from β7^hi splenic plasma cells.

Using the same non-BCR transgenic S1pr2-creERT2/tdTomato mice, we also analyzed the characteristic mutations of the V_H186.2 sequences for high-affinity antibodies to NP (Allen et al., 1988). As shown in Fig. 1 F, the frequency of the canonical affinity-improving mutation (replacement of Trp33 with Leu33; W33L⁺) was somewhat higher in tdTomato⁺β7^hi splenic and bone marrow plasma cell fractions than the tdTomato⁺β7^lo splenic fraction.

Mechanistically, two possibilities exist to explain why newly generated integrin β7^lo plasma cells cannot migrate to the bone marrow; failure to egress from secondary lymphoid organs and/or to migrate from blood into tissues, including the bone marrow parenchyma (Kometani and Kurosaki, 2015; Lindquist et al., 2019). Flow analysis (Fig. 1 G) suggested that the integrin β7^hi plasma cells are superior in their ability to egress from the spleen compared with β7^lo cells. Hence, egress into blood is a likely primary reason for their differential homing characteristics.

To determine whether the above results are valid for viral infection, we extended our analysis to an influenza virus infection model. The primary target of antibodies directed against the influenza virus is its hemagglutinin (HA) protein. After infection of S1pr2-creERT2/tdTomato mice with influenza virus (strain PR8) and treatment with tamoxifen, we analyzed tdTomato⁺HA-specific CD138^hiTACI^hi plasma cells in mediastinal LNs (MLNs) and bone marrow (Fig. 1 H). As shown in Fig. 1 I, like with NP-CGG immunization, recently generated plasma cells in MLN showed a bimodal expression pattern of integrin β7, whereas only β7^hi cells were present in the bone marrow. Then, we wished to compare the affinity of anti-HA antibodies between integrin β7^lo and β7^hi MLN plasma cell fractions. To do so, we produced monoclonal antibodies from single-cell sorted tdTomato⁺HA-specific cells from these two fractions. Different from NP-CGG, the PR8 HA protein has five major epitopes in the head region (Angeletti et al., 2017). Hence, we sought to compare affinities of the GC-experienced antibodies derived from the same clone, which presumably recognize the same epitope. As demonstrated in Fig. S3 E, slightly higher affinity antibodies are more accumulated in the integrin β7^hi MLN plasma cell fraction in one clone (HV1–50 HJ4), but not in the other clones. Collectively, influenza infection also generates both integrin β7^lo and β7^hi plasma cells in draining LNs, the latter of which are prone to migrate to bone marrow, as was also observed in the NP-CGG immunization system.

We next sought to address two related questions: first, what is the key upstream driver that confers bone marrow trafficking ability on integrin β7^hi splenic plasma cells; second, what are the key effector molecules that make integrin β7^hi plasma cells egress and/or migrate to the bone marrow. As to the first question, given that integrin β7 expression is defective in B cells derived from mice with CD19-Cre–mediated Klf2 deletion (Winkelmann et al., 2011; Hart et al., 2011), we wondered whether KLF2 might also be an upstream candidate for regulating the expression of integrin β7. Indeed, our gene set enrichment analysis (GSEA) demonstrated that KLF2 targets were enriched in the tdTomato⁺NP⁺β7^hi splenic plasma cell fraction compared with their β7^lo counterparts (Fig. 2 A). Klf2 expression itself was also higher in the integrin β7^hi splenic plasma cell fraction (Fig. 2 A).

Based on the above results, we hypothesized that the differences in KLF2 expression between β7^lo and β7^hi plasma cells might determine their retention versus bone marrow homing. To test this hypothesis, we transferred Klf2^+/+ R26-creERT2 B1–8^hi or Klf2^f/f R26-creERT2 B1–8^hi B cells into WT recipient mice, which were then immunized with NP-CGG/alum and treated with tamoxifen (Fig. 2 B). In this experimental setting, about 90% of the donor-derived plasma cells in spleen was derived from pre-GC reactions (Fig. S2 D). Klf2 mRNA expression in splenic Klf2^f/f R26-creERT2 plasma cells was almost completely abolished (Fig. 2 C). Total numbers of donor-derived Klf2^f/f R26-creERT2 plasma cells in the spleen were somewhat increased, compared with Klf2^+/+ R26-creERT2 plasma cells, whereas these mutant plasma cells were barely detectable in blood and bone marrow (Fig. 2 D), demonstrating that KLF2 participates in the egress of newly generated plasma cells from secondary lymphoid tissues. As expected, Klf2 expression appeared to correlate well with cell surface expression of integrin β7 and CD11b; in Klf2-deficient splenic plasma cells, fractions of integrin β7^hiCD11b^hi and β7^hiCD11b^lo plasma cells were decreased completely or to a large extent, respectively (Fig. 2 E). We also carried out haploinsufficiency experiments, demonstrating that loss of one copy of Klf2 affected egress of newly generated plasma cells to some extent (Fig. 2, F and G); in a competitive setting, Klf2 haploinsufficient plasma cells accumulated more than WT cells in the spleen. Together, the difference in KLF2 expression between β7^hi and β7^lo populations in secondary lymphoid organs is likely to explain why integrin β7^hi plasma cells can preferentially egress.

Mechanistically, it is possible that the defect in splenic egress of KLF2-deficient plasma cells might be caused by their mis-localization in the spleen. To test this possibility, we examined the plasma cell distribution within the spleen by employing the above adoptive transfer system. Immunofluorescence analysis revealed that the distribution of Klf2-deficient plasma cells in marginal zone (MZ) bridging channels and red pulp was similar to that of Klf2-sufficient cells (Fig. S3 F). Consistently, surface expression of CXCR4, which is required for plasma cell positioning in splenic red pulp (Hargreaves et al., 2001; Good-Jacobson et al., 2015), was comparable between Klf2-sufficient and -deficient plasma cells (Fig. S3 G). Additionally, expression levels of CXCR5 were also similar between the two cell types (Fig. S3 G).

Next, to look for the functional target(s) downstream of KLF2 involved in egress of newly generated plasma cells, we carried out RNA-seq analysis of mRNA from WT and Klf2-deficient splenic plasma cells by employing transfer experiments with Klf2^+/+ or Klf2^f/f R26-creERT2 B1–8^hi B cells (Fig. 2 B). WT control splenic plasma cells were further divided into integrin β7^hi and β7^lo populations. PCA showed, as expected, that total WT plasma cells were intermediate between WT β7^hi and β7^lo cells. Furthermore, Klf2-deficient cells were clustered most distantly from the WT β7^hi population (Fig. 3 A). Comparison between WT β7^hi and Klf2-deficient plasma cells identified 166 differentially expressed genes (DEGs) (Fig. S4). We further analyzed the expression of genes that have been reported to be relevant to plasma cell biology. Particularly, DEGs upregulated in the WT β7^hi population were those associated with migration and adhesion, including Itgb7, Itgam (CD11b), Sell (CD62L), or S1pr1 (Fig. 3 B and Fig. S4). DEGs downregulated in WT β7^hi population include Ccr5, Gpr155, or Il6st (Fig. S4). Among these genes, in terms of egress, we first focused on S1PR1 in Klf2-deficient plasma cells because of the following two lines of evidence. First, consistent with a previous report (Kabashima et al., 2006), FTY720 treatment blocked the egress of newly generated plasma cells (Fig. S5, A and B). Second, correlated with failure of integrin β7^lo splenic plasma cells to egress, expression of S1PR1 in these cells by flow as well as RNA analysis was almost similar to that in Klf2-deficient plasma cells (Fig. 3 C).

Although the differences in the expression levels of S1PR1 between β7^hi, β7^lo splenic plasma cells, and Klf2-deficient plasma cells were modest (Fig. 3 C), we could demonstrate that such differences in S1PR1 expression have significant consequences for plasma cell egress. We performed co-transfer experiments with B1–8^hi R26-creERT2 S1pr1^+/+ and B1–8^hi R26-creERT2 S1pr1^{fl/fl or fl/+} B cells (Fig. S5 C and Fig. 3 D), in which tamoxifen treatment resulted in decrease in S1PR1 expression on plasma cells differentiated from the latter B cells (Fig. 3 E). As expected, S1pr1^fl/fl plasma cells were entirely absent from the blood or bone marrow (Fig. S5 D), confirming the essential role of this receptor in plasma cell egress. Notably, in this competitive setting, S1pr1^+/+ plasma cells were dominate over S1pr1^fl/+ haploinsufficient plasma cells in terms of their egress into blood and migration into bone marrow (Fig. 3 F), suggesting that lower S1PR1 expression in integrin β7^lo plasma cells or Klf2-deficient plasma cells could account at least in part for their failure to egress.

To test the idea that S1PR1 acts as a functional downstream effector molecule of KLF2, we overexpressed S1PR1 in a Klf2 knockout background; both overexpression of S1PR1 and deletion of Klf2 were manipulated in a simultaneous inducible manner (Fig. 3 G). As seen in Fig. 3 H, overexpression of S1PR1 in Klf2-deficient plasma cells allowed them to appear in the blood but barely in the bone marrow; the augmented egress is probably due to overexpression of S1PR1, compared with Klf2-sufficient plasma cells (Fig. S5 E). Hence, defective egress by loss of KLF2 is likely due to a low level of S1PR1.

As a control, we also overexpressed KLF2, instead of S1PR1, in a Klf2-deficient background, manifesting reduced frequencies of plasma cells in the spleen but increased frequencies in the blood (Fig. 3 H). It is possible that high levels of KLF2 might interfere with B cell proliferation and/or survival of plasma cells in the spleen. Rather, we favor the explanation that KLF2 overexpression promotes the egress of splenic plasma cells into the blood stream.

The above results raised the question of why Klf2-deficient plasma cells with overexpression of SIPR1 cannot migrate efficiently to the bone marrow. For this, we considered two non-mutually exclusive possibilities. First, circulating plasma cells might still require KLF2 for homing to bone marrow parenchyma. Second, in a physiological context, since the S1P level is high in blood, functional desensitization of the S1PR1 by S1P occurs (Lo et al., 2005), being required to allow circulating plasma cells to migrate to bone marrow parenchyma.

To test the former possibility, we utilized i.v. transfer experiments to bypass the egress process (Fig. 4 A). To distinguish sinusoid and parenchyma localization in the bone marrow, we injected anti-CD45 antibody i.v. 2 min before analysis to label only sinusoidal plasma cells. As shown in Fig. 4 B, both WT- and Klf2-deficient plasma cells moved into the bone marrow parenchyma and survived there for 30 days, suggesting that KLF2 is dispensable for plasma cell entry into and long-term survival in the bone marrow. Hence, the second possibility is more likely. In the case of lymphocytes, it was already reported that GRK2-dependent S1PR1 desensitization is required for their entry into lymphoid tissue from blood circulation (Arnon et al., 2011). Our two lines of observations strongly suggest that this type of regulation also takes place in plasma cells. First, in contrast to splenic plasma cells, the S1PR1 expression on plasma cells in the blood was downregulated in a physiological setting (Fig. 4 C). Second, Klf2-deficient plasma cells with retrovirally induced S1PR1 showed a very high level of S1PR1 expression even in the circulating blood (Fig. S5 E).

Low levels of integrin β7 and CD11b expression in Klf2-deficient plasma cells prompted us to examine effects of such adhesion molecules on bone marrow tropism of newly generated plasma cells. To genetically examine the necessity of integrin β7, we established integrin Itgb7^fl/fl R26-creERT2 B1–8^hi B cells, transferred them into recipient mice, immunized with NP-CGG/alum, and then treated with tamoxifen (Fig. 5 A). Although integrin β7 expression was almost completely abolished in donor plasma cells (Fig. 5 B), migration to the bone marrow was not affected by loss of integrin β7 (Fig. 5 C). Thus, although surface expression of integrin β7^hi specifically marks plasma cells that are prone to migrate to the bone marrow, integrin β7 by itself is not required for such trafficking.

Similarly, we considered the possible role of CD11b in plasma cell homing. For this, B1–8^germ B cells were transduced with retroviruses containing small hairpin (sh) RNA for control and CD11b, cultured, transferred into congenically marked recipient mice, and immunized with NP-CGG/alum (Fig. 5 D). Assessed by flow analysis for GFP⁺ virus–transduced cells, downregulation of CD11b was not complete, but about 60% reduction was achieved in the case of shCD11b (Fig. 5 E). In contrast to the shControl, homing to bone marrow of donor plasma cells with shCD11b was decreased (Fig. 5 F). This was mainly due to defective egress from spleen (Fig. 5 F). We next overexpressed CD11b in Klf2-deficient plasma cells (Fig. S5, F and G). Unexpectedly, this did not enable their egress from the spleen (Fig. S5 H). Thus, we conclude that while KLF2-mediated CD11b expression on plasma cells is necessary, it is not sufficient for their egress from the spleen.

To determine whether the loci of key effector molecules are a direct target of KLF2, we performed chromatin immunoprecipitation (ChIP)–quantitative PCR (qPCR) analysis. Due to the low expression levels of KLF2 in IgG1 plasma cells (∼10-fold lower than in CD8⁺ T cells; data not shown), we were unable to detect significant binding in native plasma cells. Therefore, we overexpressed FLAG-KLF2 in plasma cells by retrovirally transducing FLAG-tagged KLF2 into in vitro–generated Klf2-deficient plasma cells and then conducted the experiments. Under these conditions, as shown in Fig. 5 G, FLAG-KLF2 exhibited significant binding to the Itgb7 and S1pr1 loci, but not the Itgam locus.

The evidence that Klf2-deficeint plasma cells cannot migrate to bone marrow, together with barely detectable KLF2 expression in plasma cells remaining in the spleen, prompted us to consider that the inducible Klf2 knockout is a suitable experimental systems to address the biological significance of plasma cell migration and subsequent bone marrow residency. Hence, we generated chimeric mice, in which bone marrow from Klf2^+/+S1pr2-creERT2 or Klf2^f/fS1pr2-creERT2 were mixed with bone marrow from B cell–deficient μMT mice (Fig. 6 A). The mice were vaccinated with inactivated whole PR8 influenza virus, treated with tamoxifen, and tested for the serum anti-HA antibody response. Comparable levels of anti-HA antibodies were initially elicited in both groups of mice. However, in sharp contrast to control mice, the anti-HA antibody titer decreased over time in GC-specific Klf2-deficient mice (Fig. 6 B). As a result, sera collected from control mice conferred protection for naïve mice from infection with a lethal dose of influenza virus, whereas sera from GC-specific Klf2-deficient mice failed to do so (Fig. 6 C). Taken together, KLF2-dependent plasma cell egress is likely to be required for sustained and protective antibody response against viral reinfection.

Clarifying the key determinants of plasma cell persistence is important to regulate antibody longevity in vaccine and autoimmune settings. Since LLPCs are lodged in a location distinct from their site of generation, two conceptual models (induction site and effector tissue models) have been proposed to determine the persistence of plasma cells (Tarlinton et al., 2023). Here, we found that two distinct plasma cell populations (integrin β7^hi and β7^lo) are generated in secondary lymphoid tissues and that differential expression of KLF2 in these two populations is one of the key factors to determine their retention in secondary lymphoid organs versus homing to the bone marrow. Since bone marrow maintains more plasma cells compared with those in the secondary lymphoid organs, the KLF2 expression level in plasma cells at the induction site is one of the key variables that instruct antibody durability by regulating their migration program.

Among many targets of KLF2, S1PR1 is key to explain defective egress of KLF2-deficient plasma cells. Although the pathway by which plasma cells exit the spleen is still not well defined, it is likely that the cells need to reach red pulp sinusoids to enter blood circulation; in LNs, they are placed near lymphatic sinuses in medullary cords (Fooksman et al., 2010). It seems that CXCR4 normally contributes to retention of plasma cells in spleen and LNs. Supporting this idea, CXCR4-deficient plasma cells were mis-localized in the splenic red pulp and overrepresented in the blood (Hargreaves et al., 2001). In this regard, one of the functions of S1PR1 may be to antagonize the CXCR4-mediated retention signal in secondary lymphoid organs. Consistent with previous results (Kabashima et al., 2006), we also found that CXCR4 expression on integrin β7^hi blood plasma cells was downregulated, compared with their splenic counterparts (data not shown). Thus, we would like to propose that integrin β7^hi splenic plasma cells undergo modulation of CXCR4, and that the resulting reduced CXCR4 expression together with high S1PR1 expression contributes to shifting the balance between retention and egress signals. Consistent with this proposal, gain-of-function mutations in CXCR4 affecting receptor desensitization were reported to result in barely detectable antigen-specific plasma cells in the bone marrow, despite their increased numbers in the spleen (Biajoux et al., 2016). Although CXCR4 is partially downregulated, we anticipate that this decreased expression level is still sufficient for homing of plasma cells from blood to bone marrow (Underhill et al., 2003; Wehrli et al., 2001; Hauser et al., 2002).

In our search for other effectors beyond S1PR1 downstream of KLF2, we also found that CD11b contributes to the egress of newly generated plasma cells from spleen. However, CD11b does not appear to be a direct target of KLF2 and is not sufficient on its own for plasma cell egress. Hence, the regulation or function of CD11b expression in plasma cells require further investigation. CD11b is one of the α integrin subunits pairing with CD18 (integrin β2) (Ahn et al., 2010). Our data are entirely consistent with previous observations that, in the absence of β2-containing integrins, plasma cells accumulate in the medullary cords but are unable to exit the peripheral LNs (Pabst et al., 2005). Since both CD11b/CD18 and CD11a (integrin αL)/CD18 integrins bind to ICAM-1, the simple interpretation of our data is that binding to ICAM-1 is necessary, but binding through CD11a/CD18 integrin is not sufficient for egress of plasma cells from the spleen. Another possibility is suggested from the observation that CD11b/CD18 integrin specifically participates in the transepithelial migration of neutrophils by binding to members of the junctional adhesion molecule (JAM) family (Zen et al., 2004; Nourshargh and Alon, 2014). JAM proteins are uniquely located at tight junctions between adjacent endothelial cells and serve as ligands for CD11b/CD18 integrin to promote transendothelial migration of the cells. Assuming that a similar mechanism also operates in plasma cell egress from spleen, the adhesive interaction of the CD11b/CD18 integrin with JAM proteins may facilitate egress of integrin β7^hi plasma cells in concert with S1PR1-mediated signaling.

The MZ is rich in S1P, and MZ B cells are attracted into this compartment in a S1PR1-dependent manner (Arnon et al., 2013; Cinamon et al., 2004, 2008). However, S1PR1 on MZ B cells is then downregulated within minutes in a GRK2-dependent manner, thereby making the B cells lose much of their attraction to this location (Arnon et al., 2011). Instead, CXCL13-mediated attraction becomes dominant for B cells to move into the follicle. The mechanism of plasma cell homing from circulating blood to bone marrow parenchyma is likely to utilize a similar mechanism. Since S1P is high in blood, expression of S1PR1 is downregulated in circulating plasma cells. Due to S1PR1 desensitization, CXCL12-CXCR4–mediated attraction to the bone marrow parenchyma likely becomes dominant. However, different from the shuttling of MZ B cells between the MZ and follicle, after plasma cells home to the bone marrow, they are retained there. To do so, they might undergo transcriptional downregulation of S1PR1 and/or upregulation of CD69 in the bone marrow.

How is the expression of KLF2 regulated in secondary lymphoid tissues, leading to the generation of integrin β7^hi and β7^lo plasma cells? KLF2-reporter experiments at plasma cell generation sites have provided some indications; in the mesenteric LN, IgA⁺, but not IgG⁺, plasma cells express high levels of KLF2 (Wittner et al., 2022). Given that retinoic acid produced by dendritic and stromal cells in mesenteric LNs induces IgA class switching and integrin α₄β₇ (one of the downstream targets of KLF2) on IgA⁺ plasma cells (Hammerschmidt et al., 2008; Molenaar et al., 2009), the retinoic acid at the plasma cell induction site is likely one of the crucial factors to regulate KLF2 expression in the case of the mesenteric LNs. Assuming the existence of heterogeneity in microenvironments along the route from the plasma cell generation site to the exit site in secondary lymphoid organs (Zehentmeier et al., 2014) in the NP-CGG immunization setting, such cellular and molecular heterogeneity might affect the expression levels of KLF2, thereby generating integrin β7^hi and β7^lo plasma cells. Apart from KLF2, given that some of the β7^hi plasma cells remain in the spleen, it is conceivable that other bottom neck factors, in conjunction with KLF2, might facilitate the migration of small fractions of the generated plasma cells to the bone marrow.

Although LLPCs are well-known to inhabit the bone marrow, many other tissues hosting LLPCs have recently been discovered, including mucosal lamina propria, skin, meninges, and visceral adipose tissues (Tarlinton et al., 2023). Subsequent transcriptomic analysis indicates the existence of distinct signatures between distinct tissues, suggesting that signals from the tissue of residency might outweigh the imprinting from the induction site (Tellier et al., 2023). However, a granular analysis of the bone marrow LLPC population revealed a heterogeneity that correlates with the origin of the cells, in that each LLPC retains the expression of genes from its induction site (Tellier et al., 2023). For instance, among the bone marrow–resident LLPCs, IgA and IgG LLPCs have distinct transcriptomes. Hence, it is likely that the transcriptional differences are induced first at the induction site and then maintained. Next, after reaching the effector tissues, these plasma cells acquire additional levels of complexity superimposed on the predetermined transcriptome. Thus, probably several key extrinsic and intrinsic signals exist at both layers, and further studies are warranted to understand how these signals regulate each other in a spatiotemporal manner.

C57BL/6 mice were purchased from CLEA Japan. All of the following mice were backcrossed to B6 or congenic B6 CD45.1⁺ mouse strains: B1–8^germ (Shinnakasu et al., 2016), B1–8^hi (Shih et al., 2002), S1pr2-creERT2 (Shinnakasu et al., 2016), R26-LSL-tdTomato (Ai14) (Madisen et al., 2010), Klf2^fl/fl (Weinreich et al., 2009), S1pr1^fl/fl (Allende et al., 2003), and µMT (Kitamura and Rajewsky, 1992). R26-creERT2 mice were obtained from Taconic Farms. Itgb7^fl/fl mice were generated by the standard CRISPR-Cas9 approach by inserting two loxP sites into intronic regions upstream and downstream of exon 4. All mice were bred and maintained under specific pathogen–free conditions, and all animal experiments were performed under the institutional guidelines of Osaka University.

Single-cell suspension from the spleen, LNs, bone marrow (tibia and femur), and peripheral blood were depleted of RBCs by RBC Lysis Buffer (Thermo Fisher Scientific), blocked with anti-CD16/32 (2.4G2; BD Bioscience), and then stained with the following mAbs for 30 min at 4°C in PBS supplemented with 0.5% BSA and 2 mM EDTA: from BD Biosciences, CD11b (M1/70, 563553), CD11c (HL3, 561119), CD138 (281-2, 740880), CD45.1 (A20, 565212), CXCR4 (2B11, 558644), IgG1 (A85-1, 746811; 740121; 553443), streptavidin (564876; 557598), and TACI (8F10, 742840); from Bio- Rad, IgG2c (STAR135F); from Biolegend, CCR7 (4B12, 120107), CD11b (M1/70, 101212; 101208), CD138 (281-2, 142508), CD44 (IM7, 103030), CD45.2 (104, 109808), CXCR5 (L138D7, 145505), IgG2b (RMG2b-1, 406706), integrin α4 (R1-2, 103621), integrin α4β7 (DATK32, 120607), integrin αL (M17/4, 101119), integrin αLβ2 (H155–78, 141009), integrin β1 (HMβ1-1, 102215), integrin β2 (M18/2, 101414), integrin β7 (FIB504, 321208; 321209; 321242), and TACI (8F10, 133404); from eBioscience, CD69 (H1.2F3, 17–0691-82); from Invitrogen, CD11b (M1/70, 47–0112-82; 25–0112-81), CD45.1 (A20, 17–0453-82; 11–0453-85; 12–0453-83), CD62L (MEL-14, 17–0621-82), IgM (II/41, 17–5790-82; 47–5790-82), and streptavidin (17–4317-82); from Jackson ImmunoResearch, Rat IgG (712-065-153); from R&D systems, S1P1/EDG-1 (713412).

Staining of surface S1PR1 was performed on ice in PBS containing 2% FBS, 1 mM EDTA, and 2% normal mouse serum, as described previously (Arnon et al., 2011; Baeyens et al., 2021) with the following modifications. After incubation with mouse IgG2a (dilution 1:10; 0103-01; Southern Biotech) for 30 min, cells were sequentially incubated with rat anti-mouse S1PR1 (dilution 1:10, MAB7089; R&D Systems) for 90 min, anti-rat IgG-biotin F(ab′)2 (dilution 1:100, 2340649; Jackson ImmunoResearch) for 45 min, 2% normal rat serum for 30 min, and streptavidin-APC (dilution 1:100, 17–4317-82; eBioscience) for 30 min.

NP-binding Ig was detected with NP₁₄-BSA-Alexa Fluor 647 (Haniuda et al., 2016). Dead cells, detected with propidium iodide or Fixable Viability Dye (eBioscience), were gated out in all FCM experiments. In some experiments, plasma cells were enriched before staining of surface molecules by incubating first with anti-CD138 (281-2; BD Bioscience), followed by magnetic enrichment with anti-rat IgG MicroBeads (Miltenyi Biotech). Stained cells were analyzed using FACSymphony A3, FACSCantoII (BD Biosciences), or Attune NxT (Thermo Fisher Scientific), or sorted on FACSAriaII or FACSymphony S6 (BD Biosciences) instruments. The data were analyzed using Flowjo (Tree Star).

For the LEGENDScreen (catalog number 700009; BioLegend) assay, SPL B1–8^germ CD138⁺ cells (CD45.1/2) or BM B1–8^germ CD138⁺ cells (CD45.1/1) were enriched from spleen or bone marrow, respectively, of adoptive transfer recipients. Cells were pooled, stained as described above, and transferred into each well of LEGENDScreen plates, which were reconstituted according to the manufacturer’s instructions. Plates were incubated with 30 min at 4°C. Cells were washed, fixed, and analyzed on FACSAria II.

Mice were immunized i.p. with 30 μg of NP-CGG in alum (Inoue et al., 2021). To induce the deletion of the loxP-flanked STOP cassette by Cre-mediated recombination, 2 mg tamoxifen (Sigma-Aldrich) in sunflower seed oil (Sigma-Aldrich) was orally administered to mice once per day for 2 or 3 days. To label bone marrow sinusoidal cells, 1 μg of PE-conjugated anti-CD45 (30-F11, 103106) was injected i.v. 2 min before sacrifice. For FTY720 treatment, the mice were injected i.p. with 3.0 mg/kg of FTY720 (Sigma-Aldrich) dissolved in 0.9% NaCl (saline).

B cells were purified from spleen of B1–8^germ or B1–8^hi mice by magnetic cell depletion with CD43 MicroBeads (Miltenyi Biotech) or with Streptavidin MircoBeads (Miltenyi Biotech) and biotinylated antibodies against CD43 (S7; BD), CD4 (GK1.5; BioLegend), CD8a (53-6.7; Invitrogen), CD11b (M1/70; eBioscience), Gr-1 (RB6–8C5; eBioscience), NK1.1 (PK136; Invitrogen), and Ter119 (TER-119; Invitrogen). The frequency of NP⁺ cells among B cells was determined by FCM. The naïve B cells containing 2.5 × 10⁴ NP⁺ B cells per mouse were transferred i.v. into B6 mice, which were then immunized i.p. with NP-CGG in alum on the next day.

For plasma cell transfer, B1–8^hi donor-derived CD138⁺ cells were enriched by magnetic cell depletion with Streptavidin MircoBeads (Miltenyi Biotech) and biotinylated antibodies against CD4 (GK1.5; BioLegend), CD8a (53-6.7; Invitrogen), Igκ (187.1; BD Bioscience), and Ter119 (TER-119; Invitrogen). In some experiments, enriched B1–8^hi CD138⁺ cells were further purified into integrin β7^hi or β7^lo populations for adoptive transfer.

Mixed bone marrow chimeras were generated as follows: C57BL6 mice were lethally irradiated (8.5 Gy) and reconstituted with a mixed inoculation of 80% µMT and 20% Klf2^+/+S1pr2-creERT2 or 20% Klf2^fl/flS1pr2-creERT2 bone marrow cells for 8 wk. The chimeras were then vaccinated i.p. with 50 µg of inactivated whole PR8 influenza virus and were administered with tamoxifen at day 7 to day 9. Serum was collected at day 10, 20, 40, and 60. Naïve C57BL6 mice that were transferred i.p. with 200 µl of sera from naïve mice or vaccinated chimeras were inoculated intranasally with 0.1× LD50 or 5× LD50 H1N1 A/Puerto Rico/8/1934 (PR8) virus, respectively.

To inducibly express KLF2, S1pr1, or CD11b in Klf2-deficient B cells in vivo, Klf2^fl/fl R26-creERT2 B1–8^hi (CD45.1⁺) mice were immunized i.p. with 50 µg of NP-Ficoll in PBS. 6 h later, splenic B cells were purified and cultured with 2 µg/ml of anti-CD40 (clone 3/23; BD Biosciences) for 18 h. Cells were then retrovirally transduced with an inducible KLF2 or S1pr1 expression cassette by spin infection (800 g, 90 min, 25°C) with polybrene (8 µg/ml; Millipore) and retroviral supernatant produced in Plat-E cells. The retroviral vector was constructed by inserting a flip-excision (FLEX) Cre-switch cassette, mouse Klf2, S1pr1, or Itgam cDNA, and IRES-eGFP into the pMYs vector. 3 h after infection, 0.5–1 × 10⁶ B cells were transferred i.v. into C57BL6 mice, which were immunized i.p. with 30 µg of NP-CGG/alum on day 0, followed by tamoxifen treatment on days 3, 4, and 5 to induce Klf2 deletion and Klf2, S1pr1, or CD11b expression.

To generate plasma cells expressing FLAG-tagged KLF2, B cells from Klf2^fl/fl R26-creERT2 mice were cultured using iGB method as described previously (Nojima et al., 2011; Haniuda et al., 2016). Briefly, B cells were cultured on 40LB feeder cells with IL-4 (1 ng/ml; PeproTech). On day 2, cells were infected with either pMYs-GFP-FLEX-empty or pMYs-GFP-FLEX-3xFLAG-KLF2 retrovirus. On day 3, cells were harvested and transferred to new 40LB feeder layers supplemented with IL-21 (10 ng/ml; PeproTech) for an additional 3 days. On day 6, iGB cells were isolated and further cultured on fresh 40LB feeder layers with IL-21 (10 ng/ml) and 4-OHT (250 ng/ml; Sigma-Aldrich) to induce the expression of 3xFLAG-KLF2. Finally, on day 8, cells were harvested, and CD138⁺IgG1⁺GFP⁺ cells were sorted for ChIP-PCR analysis.

To reduce CD11b expression, B cells from B1–8^germ (CD45.1⁺) mice were transduced as described above, with retroviruses containing a sh RNA that targets CD11b. Transduced B cells were transferred i.v. into C57BL6 mice, followed by immunization with NP-CGG/alum. The retroviral vector was constructed by inserting the CD11b-targeting sequence (5′-GGA​GCC​TTC​CTG​TAT​ACA​TCG-3′) into the pSIREN-GFP vector.

NP-specific IgG1⁺ CD138^hi TACI^hi cells in bone marrow or spleen from mice immunized with NP-CGG were single-cell sorted. cDNA was synthesized, and Igγ/Igλ sequences were analyzed as described previously (Inoue et al., 2021). For the analysis of clonal similarity between the cell types, CDR3 of each cell was determined by IMGT/V-Quest, the counts of each CDR3 were calculated using VDJtools, and the correlation coefficient of CDR3 between the cell types was calculated by CORREL function in Excel.

For the analysis of HA-specific plasma cells, HA⁺IgG⁺ CD138^hi TACI^hi cells in the spleen from mice infected with influenza PR8 virus were single-cell sorted. BCR cloning and mAb expression were performed as described previously (Inoue et al., 2023) with the following modifications. PCR-amplified Igγ/Igκ transcripts were cloned into the mouse Igγ2c/Igκ-expression vector using the SLiCE method (Motohashi, 2015). mAbs were expressed using the Expi293 Expression System (Thermo Fisher Scientific) and purified from the culture supernatants of Expi293F cells by Protein G Mag Sepharose (Cytiva) The affinity of mAb against HA was determined with the OctetRED96e system (ForteBio) as described previously (Shinnakasu et al., 2021).

For Digital RNA-seq, one hundred IgG1⁺CD138^hiTACI^hi cells derived from B1–8^germS1pr2-creERT2/tdTomato donor B cells were collected in SingleCellProtect Single Cell Stabilizing Solution (Avidin) and frozen in liquid nitrogen. Digital RNA-seq was performed as previously described (Jin et al., 2023) using DNA molecular barcodes (Shiroguchi et al., 2012; Ogawa et al., 2017). Briefly, the libraries were generated and sequenced with different sample indexes on the MiSeq platform (150 cycles; Illumina kit). Sequencing data were mapped to the mouse genome (mm10 assembly from the UCSC Genome Browser; annotation refFlat from the UCSC Genome Browser) using STAR ver.2.7.3a. Sequencing datasets have been deposited in the Gene Expression Omnibus under accession code GSE266284.

For bulk RNA-seq, total RNA was extracted from sorted plasma cells using a RNeasy Micro Kit (Qiagen). Full-length DNA was generated using a SMART-Seq HT Kit (Takara Bio) according to the manufacturer’s instructions. An Illumina library was prepared using a NexteraXT DNA Library Preparation Kit (Illumina) according to SMARTer kit instructions. Sequencing was performed on an Illumina NovaSeq 6000 sequencer (Illumina) in the 100-base paired-end mode. After adapter trimming by Trimmomatic, sequenced reads were mapped to the mouse reference genome sequences (mm10) using TopHat version 2.1.1. The fragments per kilobase of exons per million mapped fragments were calculated using Cuffnorm version 2.2.1. The raw data have been deposited in the Gene Expression Omnibus database under accession code GSE266441.

For differential gene expression analyses and PCA, tag counts obtained by FeatureCounts (Subread package) were analyzed using DESeq2 package in R. Hierarchical clustering was performed by hclust function (ward.D method), followed by heatmap depiction using the heatmap.2 function in R package gplots. Gene set enrichment analysis was performed with software from the Broad Institute (Subramanian et al., 2005) with the following settings: permutation type = gene_set, scoring = weighted, metric = Signal2Noise, and collapse = No_Collapse. Gene set of “KLF2_TARGETS_UP” was generated from GSE160732 (Wittner et al., 2022), defined as the significantly upregulated (false discovery rate < 0.05) genes in WTcre compared with KLF2-cKO.

Plasma cells generated in vitro were cross-linked with 1% formaldehyde for 30 min at room temperature. Following nuclear extraction, the cross-linked DNA was fragmented by sonication using Digital Sonifier (Branson). The lysate was incubated overnight at 4°C with 100 μl Dynabeads anti-mouse IgG magnetic beads (Thermo Fisher Scientific) that had been preincubated with 5 μg anti-FLAG antibodies (M2; Sigma-Aldrich). After washing, samples were eluted, reverse cross-linked at 65°C for 24 h, purified using the ChIP DNA Clean & Concentrator (Zymo Research), and subjected to qPCR analysis. The results were normalized to the input DNA.

Real-time qPCR was performed as described previously (Ise et al., 2018). The following primers were used for qPCR analysis: Klf2 forward, 5′-GCG​TAC​ACA​CAC​AGG​TGA​GA-3′; Klf2 reverse, 5′-ACA​TGT​GTC​GCT​TCA​TGT​GC-3′; S1pr1 forward, 5′-TTG​GAG​TGC​ACC​GCT​GA-3′; S1pr1 reverse, 5′-AAC​AGC​AGC​CTC​GCT​CA-3′; and Actb forward, 5′-GCC​TTC​CTT​CTT​GGG​TAT​GGA-3′; Actb reverse, 5′-ACG​GAT​GTC​AAC​GTC​ACA​CT-3′.

The following primers were used for ChIP-qPCR analysis: Itgam ChIP forward, 5′-CAG​ATG​TGG​CAG​GAA​GGG​AG-3′; Itgam ChIP reverse, 5′-ACC​TGG​CAA​GTC​AAC​TTA​GGA-3′; Itgb7 ChIP forward, 5′-CTC​TCC​CTG​AGA​GCA​GAA​GC-3′; Itgb7 ChIP reverse, 5′-TAC​TGA​GCC​GTT​GAA​CAC​CC-3′; S1pr1 ChIP forward, 5′-AGC​ATT​AAC​CCC​TCC​CAG​TC-3′; S1pr1 ChIP reverse, 5′-CCA​CTT​TTG​CAA​GAC​GAA​GTC-3′; and Cd4 ChIP forward, 5′-CAC​AGA​GCC​GTG​TTG​ACT​CT-3′; Cd4 ChIP reverse, 5′-CGT​TGC​CCA​ACA​GAT​CCA​CT-3′.

Unfixed spleens were embedded in the Tissue-Tek O.C.T. Compound (4583; Sakura Finetek). Frozen sections of 10 μm in thickness were blocked with 0.1% globulin-free BSA (013–15104; FUJIFILM Wako) in PBS for 1 h at room temperature and then stained with the following fluorophore-conjugated antibodies (BioLegend) for 2 h at room temperature: anti-mouse CD45.1-PE (A20, 110707), anti-mouse CD138-APC (281-2, 142505), anti-mouse CD169-BV421 (3D6.112, 142421), and anti-mouse IgD-FITC (11–26c.2a, 405703). Sections were mounted with the FluorSave Reagent (34578920ML; Merck). Images were acquired at room temperature with a FV31S-SW using the plan apochromat UPLXAPO objective (20× with 0.80 NA) on the FLUOVIEW FV3000 inverted confocal microscope (EVIDENT).

The 96-well plates (Nunc MaxiSorp; Thermo Fisher Scientific) were coated with streptavidin (Funakoshi) at a concentration of 10 U/ml. After blocking with Blocking One solution (Nacalai Tesque), the plates were incubated with biotinylated rHA, followed by incubation with diluted mouse serum. After treatment with 5M urea to remove low avidity antibodies, HA-specific IgG antibodies were detected using HRP-conjugated anti-mouse IgG (Southern Biotech) with SureBlue TMB Substrate (KPL). The absorbance at 450 nm was measured with a microplate reader (MultiScan FC; Thermo Fisher Scientific). A HA-specific, recombinant mouse IgG mAb, developed in-house, was included on each plate for serum samples to convert OD values into relative antibody concentrations.

Statistical analysis was performed with a two-tailed unpaired Student’s t test, a two-tailed paired Student’s t test, a Fisher’s exact test, or one-way ANOVA followed by Tukey’s multiple comparisons test or log-rank (Mantel–Cox) test using GraphPad Prism software. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Fig. S1 shows the identification of integrin β7–expressing plasma cells in the spleen and bone marrow (related to Fig. 1, A and B). Fig. S2 shows the generation of integrin β7⁺ plasma cells through pre- or post-GC response (related to Fig. 1). Fig. S3 shows the survival, repertoire, and localization of antigen-specific plasma cells (related to Figs. 1 and 2). Fig. S4 shows the heatmap analysis of RNA-seq data (related to Fig. 3, A and B). Fig. S5 shows the role of S1pr1 or CD11b in plasma cell migration (related to Fig. 3, Fig. 4 C, and Fig. 5, D–F).

The RNA-seq data were deposited in the Gene Expression Omnibus database under accession numbers GSE266284 and GSE266441. All other data supporting the findings of this study are available within the article and supplemental materials.

We thank Dr. Stephen Jameson (University of Minnesota, Minneapolis, MN, USA) and Dr. Masatsugu Ema (Shiga University of Medical Science, Otsu, Japan) for Klf2-floxed mice, Dr. Masaru Ishii (Osaka University, Osaka, Japan) and Dr. Richard Proia (NIH, Bethesda, MD, USA) for S1pr1-floxed mice, and Dr. Masahiro Yamamoto (Osaka University, Osaka, Japan) for generation of Itgb7-floxed mice. We thank Dr. Oliver Bannard (University of Oxford, Oxford, UK) for detection of HA-specific plasma cells and Ms. Kaori Fukuhara (RIKEN Center for Biosystems Dynamics Research, Osaka, Japan) for her technical assistance in library preparation and digital RNA-seq.

This work was supported by Nippon Foundation (to W. Ise), Otsuka Pharmaceutical Co., (to T. Kurosaki), Grand-in-Aid for JSPS Fellows (JP22KJ2102 to T. Koike), Grand-in-Aid for Young Scientists (JP23K14543 to T. Koike), JSPS KAKENHI (JP19H01028, JP22H00450, and JP22K21354 to T. Kurosaki; JP18KK0227 and JP20H03503 to W. Ise; JP18H05411 to K. Shiroguchi), the Japan Agency for Medical Research and Development (JP223fa627002 to W. Ise and JP243fa627001 to T. Inoue), the Takeda Science Foundation (to T. Koike), the Uehara Memorial Foundation (to W. Ise), the Naito Foundation (to W. Ise), the SENSHIN Medical Research Foundation (to W. Ise), Takeda Hosho Grants for Research in Medicine (to W. Ise), and Daiichi Sankyo Foundation of Life Science (to W. Ise).

Author contributions: W. Ise: conceptualization, funding acquisition, investigation, writing—original draft, and writing—review and editing. T. Koike: formal analysis, funding acquisition, investigation, methodology, project administration, resources, validation, and writing—original draft. N. Shimada: investigation. H. Yamamoto: investigation. Y. Tai: investigation. T. Shirai: investigation. R. Kawakami: data curation, formal analysis, software, and writing—review and editing. M. Kuwabara: investigation. C. Kawai: resources. K. Shida: investigation. T. Inoue: funding acquisition, investigation, methodology, and resources. N. Hojo: investigation. K. Ichiyama: investigation. S. Sakaguchi: resources and supervision. K. Shiroguchi: investigation. K. Suzuki: investigation and writing—review and editing. T. Kurosaki: conceptualization, data curation, funding acquisition, methodology, project administration, resources, and writing—original draft, and writing—review and editing.