CXCR4 and a cell-extrinsic mechanism control immature B lymphocyte egress from bone marrow

We examined how CXCR4 inhibits BM egress by visualizing the dynamic behavior of developing B lymphocytes within calvarial BM of Rag1^GFP/+ mice by two-photon intravital microscopy (IVM) before and after disruption of CXCR4 signaling. Mice were injected i.v. with fluorescently labeled dextrans to allow distinguishing BM parenchyma from sinusoids. Approximately 99.8% of GFP⁺ cells in BM are B lineage cells (99.5% B220⁺ cells and 0.3% B220^lo CD93⁺ early B lineage progenitors), ∼0.1% are CD3e⁺ T cells (presumably recent thymic emigrants), and CD3e⁻ NK1.1⁺ cells were all found to be GFP⁻ (not depicted). GFP⁺ B lineage cells displaced throughout the 30-min imaging period (Fig. 1, A and B) at a median velocity of 2.4 µm/min (Fig. 1 C), as expected (Pereira et al., 2009). However, cells rapidly stopped their migration 1–2 min after treatment with 80 µg of the CXCR4 antagonists AMD3100 or TN14003 (Tamamura et al., 2001) i.v. or 1 h after treatment with CXCL12 blocking antibody (Fig. 1, A and B; and Video 1). The sudden arrest led to a significant reduction in the median velocity (Fig. 1 C) and mean displacement (Fig. 1 D), which were reflected in a 6–10-fold reduction of the mean motility coefficient (Fig. 1 B). CXCR4 signaling blockade also changed B cell morphology from an amoeboid/elongated form to a significantly rounded shape (Fig. 1, E and F; and Video 1). The effects of AMD3100 were reversible after 2–3 h, with most cells regaining an amoeboid shape and motility (not depicted), which is in agreement with its rapid decay rate in vivo (Hendrix et al., 2000). These data suggested an intrinsic requirement for CXCR4 for B cell migration in vivo. Indeed, B cells conditionally deficient in CXCR4 moved within BM with a significantly reduced median velocity (Fig. 1 H and Video 2) and were more rounded than WT B cells (Fig. 1 I). However, most cells were not entirely stopped (Video 2), and their mean motility coefficient was only reduced by ∼40% (Fig. 1 G). Treatment with AMD3100 (80 µg i.v.; Video 3) further reduced the velocity and mean motility coefficient of CXCR4-deficient B cells and increased their roundness (Fig. 1, G–I). The similar activities of the three CXCR4 signaling inhibitors reduced the probability that B cell motility defects were caused by off-target effects. Also, in vitro B cell migration toward the chemokine CXCL13 was unaffected by treatment with AMD3100 (not depicted). Genetic and pharmacological CXCR4 deficiency also led to the significant mobilization of developing B cell subsets from BM into blood (Fig. 1 J), as expected (Nie et al., 2004; Pereira et al., 2009). As CXCR4 is critically required for B lineage cell retention in BM, it is possible that the few GFP⁺ B lineage cells observed in calvaria BM of CXCR4 conditionally deficient mice had residual CXCR4 signaling that was sufficient for promoting BM retention and migration. Indeed, some CXCR4 conditionally deficient B lineage cells exhibited low CXCR4 surface expression (not depicted). Nevertheless, these data do not exclude the possibility that lymphocyte-extrinsic CXCR4 contributes to B lineage cell motility in BM.

Leukocyte migration is independent of integrin-mediated adhesion within lymphoid organ interstitium (Woolf et al., 2007; Lämmermann et al., 2008). In contrast, leukocytes require integrins for migration on two-dimensional surfaces (haptokinetic migration), such as migration along luminal endothelium toward the site of diapedesis (Alon and Feigelson, 2009). The cell shape changes that occurred immediately after treatment with CXCR4 inhibitors suggested that CXCR4 was involved in promoting B cell adhesion to the BM extracellular matrix and raised the possibility that B lymphocyte migration within BM interstitium was haptokinetic (i.e., integrin dependent). Consistent with this possibility, CXCR4 signaling increases α4β1 integrin affinity for its ligand VCAM-1, and both contribute to B lineage cell retention in BM (Koni et al., 2001; Leuker et al., 2001; Glodek et al., 2003; Pereira et al., 2009). To test this hypothesis, we analyzed B cell motility within BM parenchyma of Rag1^GFP/+ mice by IVM before and after i.v. administration of 200 µg anti-α4 (clone PS/2) or anti–VCAM-1 (clone M/K-2) blocking antibodies. B cell migration was largely blocked after treatment with anti-α4 or anti–VCAM-1 antibodies (Fig. 2 A and Video 4). There was a sixfold reduction of the mean motility coefficient (Fig. 2 B), reduced cell velocity (Fig. 2 C), and increased cell roundness (Fig. 2 D). The requirement for α4β1 integrin–mediated haptokinesis was intrinsic to B lineage cells as similar observations were made in Itgb1-deficient B lymphocytes (Video 5). Treatment with anti-α4 blocking antibody did not change the dynamic behavior of Itgb1-deficient B cells (not depicted). Interestingly, loosely adherent and rounded cells were seen occasionally moving short distances for a brief period (Video 6). The number of developing B cells was significantly reduced in BM, which corresponded to a significant increase in peripheral blood (PB; not depicted), as expected (Pereira et al., 2009). Electron microscopy analyzes of the BM sinusoidal endothelium demonstrated that it is highly fenestrated (Tavassoli and Yoffey, 1983). Consistent with these observations, the BM vasculature was permeable to large-molecular-mass, FITC-coupled dextrans (∼500 kD) present in circulatory fluids (Fig. 2, E and F; and Video 7), and blood plasma perfusion reached an interstitial fluid flow rate of ∼2 µm³/s (Fig. 2 G). The extensive perfusion of the BM parenchyma within minutes of injection of intravascular dyes is consistent with BM-resident cells being continuously exposed to shear stress generated by fluid flow. Indeed, considering that the femur volume is <100 µm³ (our unpublished estimates), we hypothesized that circulatory molecules access the entire femur BM parenchyma in <2 min. To test for this possibility, we injected 5 µg biotin-conjugated anti-CD19 antibody i.v., allowed it to equilibrate for 2 or 10 min, and measured the extent of binding to pro-B and pre-B cells in BM. For comparison we examined labeling of splenic follicular and marginal zone B cells of the same mouse. Our data show that the large majority of pro-B and pre-B cells are readily labeled with anti-CD19 within 2 min, and by 10 min 100% of each B cell subset is saturated with circulatory antibody (Fig. 2 H). In contrast, a significant fraction of mature follicular B cells in the spleen is poorly stained even after 10 min of exposure to biotin anti-CD19 antibody in circulation. As expected, the marginal zone B cell subset was readily stained with anti-CD19 antibodies after 2 min (Cinamon et al., 2008), but saturation was only achieved after 10 min (Fig. 2 H). These data demonstrate that circulatory molecules rapidly perfuse BM parenchyma. Whether lymphocytes access BM exit sites by forces induced upon cells, such as interstitial fluid flow, or exclusively by active migration guided by chemoattractants sensed by GPCRs intrinsically expressed on lymphocytes was not clear.

The findings that CXCR4 blockade significantly reduced B lineage cell motility in parenchyma while accelerating egress into sinusoids and PB raised an interesting and unprecedented possibility: that BM egress could be achieved in the absence of amoeboid migration toward BM exit sites. To test this hypothesis, we conditionally induced PTX expression in B lineage cells by crossing Mb1^Cre/+ mice with mice encoding the ADP-ribosyl transferase S1 subunit of PTX within the Rosa26 locus, preceded by a premature transcriptional stop codon flanked by loxP sites (for simplicity referred as Rosa26^PTX/+ mice; Regard et al., 2007). Using this strategy, >99% of B lineage cells conditionally express genes under Rosa26 promoters from the pro-B cell stage and throughout subsequent stages of development (not depicted). The number of pro-B and pre-B cells was slightly but significantly reduced in BM, which corresponded to a small, but highly significant numerical increase in blood circulation (Fig. 3, A and B). However, IgM⁺ immature B cell subsets were reduced by two- to threefold in BM and significantly increased in blood, indicating increased BM egress. We did not find evidence of bystander effects, such as PTX leakiness into neighboring cells in BM mixed chimeras (not depicted). Furthermore, similar findings were obtained with B lineage cells deficient in Gnai2 and Gnai3 (Hwang et al., 2013). PTX-expressing B lineage cells were also poorly retained within BM sinusoids (not depicted), consistent with a role for Gαi-coupled CB2 and S1PR3 in this process (Pereira et al., 2009; Donovan et al., 2010). Mature B lymphocytes were also reduced in BM (Fig. 3 A) because CXCR4 coupling to Gαi proteins is PTX sensitive, and mature B cells require CXCR4 for homing back to BM (Nie et al., 2004; Pereira et al., 2009).

To investigate whether B lineage cell motility within BM was dependent on PTX-sensitive GPCRs, we visualized the behavior of PTX-expressing B cells marked with a Rag1^GFP/+ allele in calvaria BM by IVM. We found that PTX-expressing B cells were predominantly nonmotile (Fig. 3, C and D; and Video 8) and exhibited rounded morphology (Fig. 3 E). Occasionally, we found BM pockets with some PTX-expressing B cells moving in an amoeboid manner with similar kinetics as WT cells (Video 9), but their migration was sensitive to CXCL12 blocking antibody treatment (Fig. 3 F). These findings suggest that CXCR4 can couple to PTX-insensitive G proteins in B cells and promote interstitial migration in vivo. In agreement with these observations, AMD3100 blocked the chemotaxis of WT and PTX-expressing developing B cell subsets toward a CXCL12 gradient in vitro, whereas CXCR4-deficient B cells were unable to migrate in similar conditions (not depicted). Finally, we asked whether other hematopoietic cells could leave BM independently of PTX-sensitive GPCRs. To address this question, we treated mice with 1 µg PTX i.v. for 24 h and quantified hematopoietic cell egress from BM into PB by flow cytometry. NK cells, neutrophils, and inflammatory monocytes, like B lineage cells, were efficiently mobilized from BM into PB by PTX treatment (Fig. 3 G), even though PTX-sensitive GPCRs can contribute to their export from BM (Serbina and Pamer, 2006; Walzer et al., 2007; Jenne et al., 2009; Allende et al., 2010; Eash et al., 2010; Pereira et al., 2010). Similar results were obtained by cre recombinase–mediated PTX expression in granulocytes and monocytes and in NK cells, driven by Lyz2 (Clausen et al., 1999) or by Il7r (Schlenner et al., 2010), respectively (not depicted).

The increased rate of BM B cell egress seen in CXCR4-deficient mice, or in PTX-expressing B cells, suggested that defective motility within parenchyma enforced B lineage cells to position near BM sinusoids. To address this possibility, we measured the distance between WT, CXCR4-deficient, and PTX-expressing B lineage cells and BM sinusoids and found that both CXCR4-deficient and PTX-expressing cells were significantly accumulated in proximity to sinusoids (perisinusoidal space was defined as the area that is <10 µm distal from sinusoids; Fig. 4, A and B). As CXCR4 controls the movement of B lineage cells in parenchyma and the lack of CXCR4 signaling positions cells around sinusoids (Fig. 4, A and B) and within sinusoids (Pereira et al., 2009), we hypothesized that B lineage cells moving within perisinusoidal niches had reduced motility. Thus, we analyzed the dynamic behavior of B lineage cells that moved within 10-µm distance to sinusoids and compared them with cells moving in parenchymal areas of the same imaging volume. Even though there was a trend toward reduced median velocity that did not reach statistical significance (Fig. 4 C), their motility coefficient was reduced by 35% (Fig. 4 D), and cells within perisinusoidal niches were significantly less amoeboid (Fig. 4 E). B cells were only rarely observed migrating across sinusoidal endothelium before or after CXCR4 blockade, or in B cells genetically deficient in CXCR4 or in Gαi protein–dependent receptors. This was not surprising because the numerical increase of Gαi protein– or CXCR4-deficient B cell subsets in blood is ∼10–100-fold lower than the total number of B lineage cells remaining in BM (Fig. 3, A and B; and not depicted, respectively). Furthermore, the relatively small number of B lineage cells visualized and the limited volume of imaging fields make it difficult to capture B cells transmigrating across sinusoidal endothelial barriers. However, in ∼18 h of IVM of untreated WT (Rag1^GFP/+) mice and in 3 h of IVM of PTX-expressing B cells, we were able to observe reverse transmigration of 11 and 2 B lineage cells, respectively. In all examples observed, B lineage cells that exited BM parenchyma into sinusoids did not approach BM exit sites in an amoeboid manner, were poorly motile, and resided adjacent to sinusoids (Video 10). Consistent with our previous findings, we did not observe morphological differences between WT and PTX-expressing B cells in the process of exiting into sinusoids (Fig. 4, F and G; and Video 10), even though PTX-expressing B cells were rapidly displaced into circulation after entry into sinusoids, whereas WT B lineage cells were retained in this compartment, as expected (Pereira et al., 2009; Donovan et al., 2010). Interestingly, reverse transmigration occurred exclusively in areas where the sinusoidal endothelial barrier exhibited intense vascular leakage (Video 10). In some cases, WT and PTX-expressing cells deformed during reverse transmigration, which strongly suggests that reverse transmigration across BM sinusoidal endothelium is an active process that is independent of GPCR signaling.

Previous studies have shown that CXCR4 expression is highest at the pro-B cell stage and decreases gradually during murine and human B cell development (Honczarenko et al., 1999; Pereira et al., 2009). The finding that perisinusoidal B lineage cells exhibited reduced motility and increased roundness suggested that CXCR4 signaling was reduced in egress-competent cells. To investigate this possibility, we compared CXCR4 expression on immature B lymphocyte subsets before and after BM egress and found that it was reduced by approximately twofold in cells within sinusoids as compared with cells in parenchyma (Fig. 5, A and B). This reduction was evident even after treatment with AMD3100 (not depicted), suggesting that it was not mediated by CXCR4 desensitization from exposure to perisinusoidal sources of CXCL12 (Ding and Morrison, 2013). A twofold reduction in CXCR4 was sufficient to reduce immature B lymphocyte migration toward a gradient of CXCL12 established in vitro (Fig. 5 C). To test whether a twofold down-regulation of CXCR4 was sufficient to mobilize immature B cells from BM, we compared the distribution of CXCR4 heterozygous B lineage cells in direct competition with CXCR4 WT cells in BM parenchyma, sinusoids, and PB. To accomplish this, we reconstituted lethally irradiated mice with a 50% mixture of BM cells isolated from Cxcr4^+/− or Cxcr4^+/+ CD45.1 mice with Cxcr4^+/+ CD45.2 mice. As expected, Cxcr4^+/− immature B cells in BM parenchyma expressed comparable CXCR4 surface amounts as Cxcr4^+/+ B cells positioned within BM sinusoids, whereas Cxcr4^+/− immature B cells expressed approximately twofold lower CXCR4 than Cxcr4^+/+ immature B cells in BM parenchyma (Fig. 5 D). Furthermore, Cxcr4^+/− immature B cells were significantly reduced in BM parenchyma and increased in BM sinusoids and blood when compared with Cxcr4^+/+ immature B cells (Fig. 5 E). Antagonizing CXCR4 down-regulation by retroviral transduction of hematopoietic stem cells with a desensitization-deficient CXCR4 (R334X mutation that causes the WHIM syndrome in humans [Hernandez et al., 2003]) was sufficient to prevent immature B lymphocyte egress from BM parenchyma into sinusoids and into PB (Fig. 5, F and G).

In late stages of B cell development, newly generated immature B cells undergo negative selection against systemic self-antigens predominantly in BM (Goodnow et al., 2005). This process is likely promoted by immature B cell migration within BM by facilitating cellular interactions and screening for self-reactive BCRs. As B lineage cell migration in BM is controlled by CXCR4 and reduced CXCR4 signaling is essential for BM export, we hypothesized that BCR signaling intensity controlled CXCR4 expression in immature B cells. To test this hypothesis, we analyzed CXCR4 surface expression in immature B cells before and after inducing BCR stimulation. For this purpose, we isolated BM cells from WT mice (C57BL/6), MD4 BCR transgenic mice (B cells specifically recognize hen egg lysozyme [HEL]), and B1-8 BCR transgenic mice (nitrophenyl hapten–specific B cells) and induced BCR signaling with cognate antigens for various periods of time. We found that immature B cells increased CXCR4 surface expression as early as 3 h after treatment with stimulatory anti-IgM antibodies or with cognate antigens and peaked at 12 h after BCR stimulation (Fig. 6, A and B). We then asked whether self-antigen engagement was sufficient to prevent immature B cell egress from BM. We treated MD4 transgenic mice with 10 mg HEL for 6 h i.v. and measured immature B cell subsets in BM parenchyma, sinusoids, and PB. We found that HEL treatment not only significantly increased CXCR4 expression in immature B cells in vivo (Fig. 6 C), but also prevented immature B lymphocyte egress into sinusoids and PB by almost 10-fold (Fig. 6, D and E). The effects of HEL on CXCR4 expression and B cell egress blockade were strictly dependent on an HEL-specific BCR (Fig. 6 F). The peripheral reduction in immature B lymphocyte subsets was not caused by BCR signaling–induced cell death because these cells significantly accumulated in BM parenchyma (Fig. 6 E). Furthermore, similar results were obtained with MD4 immature B lymphocytes that overexpressed antiapoptotic BCL2 (Fig. 6 H). These results predicted that antigen engagement altered B lineage cell movement in BM parenchyma. To test this hypothesis, we crossed MD4 BCR transgenic mice with mice expressing cre recombinase driven by Cd19 and with mice encoding the ZsGreen fluorescent protein driven by the chicken albumin and rabbit β globin promoter within the Rosa26 locus, preceded by a premature transcriptional stop codon flanked by loxP sites (for simplicity referred as Rosa26^ZsGreen/+; Madisen et al., 2010). Using this strategy, ∼99% of ZsGreen⁺ cells were IgM⁺ B lineage cells (not depicted). We lethally irradiated C57BL/6 recipient mice and reconstituted them with BM cells taken from MD4⁺ and MD4⁻ Cd19^Cre/+ Rosa26^ZsGreen/+ mice. To measure the effect of antigen engagement mostly in immature B lymphocyte motility in BM, we analyzed the calvaria of BM chimeras by IVM 4 wk after reconstitution, a time point where ∼75% of ZsGreen⁺ B lineage cells were immature B lymphocytes, as measured by CD93 surface expression (not depicted). In agreement with these findings, the amoeboid polarity and motility of MD4⁺ B lymphocytes increased significantly after exposure to HEL, whereas the movement and morphology of MD4⁻ B lineage cells was unaffected by HEL treatments (Fig. 6, I and J). We also noted that MD4⁺ B cells were less motile and more rounded than MD4⁻ B cells (Fig. 6, I and J). However, this difference was likely caused by the very limited size of pro-B and pre-B cell compartment in MD4 transgenic mice as compared with that in WT mice (not depicted). In summary, these data demonstrate that immature B cell motility and retention within BM parenchyma are significantly increased by BCR signaling.

Recent studies provided substantial mechanistic insight into the multistep process of lymphocyte egress from secondary lymphoid organs: as B and T lymphocytes randomly move via the activity of distinct chemoattractants, lymphocytes make frequent contacts and probe lymphoid organ sinuses. The availability of the egress-promoting chemoattractant S1P, presumably at sinusoidal exit sites, and signaling through the lymphocyte-intrinsic PTX-sensitive S1PR₁ are both required for lymphocyte reverse transmigration across sinusoidal endothelium and egress (Cyster and Schwab, 2012). Here, we demonstrate that developing B cell retention in BM parenchyma was dependent on lymphocyte-intrinsic amoeboid motility, predominantly mediated by CXCR4 signaling and by α4β1–VCAM-1. However, BM egress was surprisingly independent of egress-promoting chemoattractants sensed by PTX-sensitive GPCRs. PTX treatments in vivo have been shown to mobilize B cells and hematopoietic stem/progenitor cells from BM (Papayannopoulou et al., 2003; Ueda et al., 2004). In the work presented here, we now show that cell-intrinsic PTX expression not only reduces B lineage cell movement within BM parenchyma, but also mobilizes B lineage cells, NK cells, monocytes, and granulocytes from BM into the periphery.

The immature B cell stage is characterized by a significant change in positioning within BM, with up to 50% of immature B cells localizing within BM sinusoids in an α4β1–VCAM-1– and CB2 and S1PR₃ signaling–dependent manner (Pereira et al., 2009; Donovan et al., 2010). However, immature B cell movement toward sinusoids was independent of CB2- and S1PR₃-mediated chemotaxis and only partially contributed by S1P and S1PR₁ (Allende et al., 2010; Pereira et al., 2010). Our experiments now reveal that CXCR4 down-regulation enforced B cells to distribute into perisinusoidal and intrasinusoidal compartments and was critical for immature B cell egress from BM. Furthermore, these experiments also suggest that α4β1-mediated adhesion is temporally reduced when cells are in perisinusoidal compartments, before it is increased again within sinusoids, likely through CB2-induced transactivation (Pereira et al., 2009). It is possible that B cell–intrinsic α4β1 and VCAM-1 expressed on sinusoidal endothelial cells may antagonize reverse transmigration of developing B cell subsets and prevent their egress from BM. However, conditional deletion of VCAM-1 from endothelial (and hematopoietic) cells induced the selective mobilization of immature B cell subsets, most prominently immature IgM⁺ IgD^lo cells (which are distributed between BM parenchyma and sinusoids), without significant effects measured in parenchyma-restricted pro-B and pre-B cell subsets (Koni et al., 2001). Thus, these data are in agreement with the finding that α4β1–VCAM-1 is specifically required for sinusoidal retention (Pereira et al., 2009) and disfavor a prominent role played by α4β1–VCAM-1 in impeding migration across sinusoidal endothelium.

Immature B cells egress BM by down-regulating CXCR4 by approximately twofold through a mechanism that was antagonized by antigen-induced BCR signaling, presumably after negative selection has occurred. In vitro studies using mouse primary immature B cells and a chicken B cell line reported that antigen-induced BCR signaling reduced B cell chemotaxis toward CXCL12, while having no detectable effect on CXCR4 surface expression (Guinamard et al., 1999; Brauweiler et al., 2007). Here, using two independent BCR transgenic mouse strains and WT (polyclonal) mice, we reliably measured BCR-induced CXCR4 expression in immature B cells in vitro and in vivo and a corresponding increase in B lineage interstitial motility and retention within BM parenchyma. The discrepancy between these findings could be related to an inadequacy of the in vitro systems for measuring antigen-coated immature B cell chemotaxis. However, our experiments are consistent with the observation that CXCR4 expression was increased in immature B cells in multiple murine models of lupus (Wang et al., 2009a), given that these cells encounter self-antigens during development in BM (Li et al., 2002). It will be valuable to analyze the cellular composition and biological significance of the BM niches that attract self-reactive immature B cells.

Even though the S1P pathway contributes to promote hematopoietic cell egress from BM, namely NK cells, monocyte/osteoclast precursors, eosinophils, and immature B cells (Walzer et al., 2007; Jenne et al., 2009; Allende et al., 2010; Pereira et al., 2010; Sugita et al., 2010), its relatively minor contribution is in agreement with a marginally detectable S1P gradient between BM parenchyma and sinusoids (Jenne et al., 2009). The BM sinusoids are fenestrated and allow passive flow of circulatory fluids into the parenchyma (Tavassoli and Yoffey, 1983; Mazo et al., 1998). Interestingly, the few observed examples of B lineage cell egress from BM parenchyma into sinusoids occurred exclusively in areas where vascular leakage was readily observed. This observation raises the possibility that S1P receptors promote BM egress from areas in parenchyma that are less perfused. Whether the fenestrated nature of the BM sinusoidal endothelium is a constitutive feature or a regulated process remains unknown.

The mechanical forces responsible for purging sessile B cells from BM are not entirely clear, but such mechanisms likely promote the egress of other hematopoietic cell subsets, including poorly motile red blood cells. The dependency on integrin-mediated adhesion for motility in BM demonstrates that developing B cells adopt similar mechanisms to those used by leukocytes during migration on luminal endothelial surfaces under shear stress generated by blood flow (Alon and Feigelson, 2009). The sudden morphological changes observed immediately after CXCR4 and/or α4 integrin blockade are also reminiscent of the cell shape changes seen in T cells exposed to lymph flow in lymphatic sinuses (Grigorova et al., 2009). Immature B cells likely adapted to haptokinetic migration for retention and controlled migration in shear-rich three-dimensional environments such as BM, and possibly the spleen. Consistent with this model, marginal zone B cells move between blood-bathed splenic red pulp and bloodless follicles in the white pulp and also rely on integrin-mediated adhesion for their retention in this compartment (Arnon et al., 2013). The observation that B lineage cell exit occurred from areas of intense parenchymal perfusion, and that several B cells were largely nonmotile and remained rounded during reverse transmigration, suggests that drainage of interstitial fluids back into circulation via collecting sinusoids facilitates BM export of B lineage cells and possibly other hematopoietic cell types, including red blood cells. In agreement with this model, neutrophils are also poorly motile in BM and are stationed near or at BM exit sites for several minutes before exiting into the sinusoidal lumen (Devi et al., 2013).

Whereas CXCR4-deficient pro-B and pre-B cells were easily detected in blood and spleen, as expected (Ma et al., 1999; Ueda et al., 2004), their numbers in BM were only slightly reduced when compared with immature B cell subsets. These data indicate that early B cell subsets are less efficiently mobilized from BM than immature B cells and suggest that they use other retention mechanisms. However, the majority of GFP⁺ B lineage cells (which include pro-B and pre-B cells) remained predominantly sessile, exhibited a rounded morphology, and seemed largely nonadherent to the extracellular matrix when they lacked α4β1-mediated adhesion to VCAM-1. An alternative and attractive possibility is that pro-B and pre-B cells cannot undergo efficient reverse transmigration. Electron microscopy studies of femoral marrow of rats, rabbits, and guinea pigs documented extensive evidence of intimate interactions between immature lymphocyte pseudopodia and cell processes between adjacent sinusoidal endothelial cells (Tavassoli and Yoffey, 1983). In some instances, though not infrequent, lymphocyte pseudopodia were seen extending through endothelial cells (Hudson and Yoffey, 1966). We also observed few examples of WT and PTX-expressing B cells elongating during reverse transmigration through sinusoids. These observations suggest that reverse transmigration across some areas of the sinusoidal endothelial wall requires an active process. Our study reveals that it is independent of GPCR signaling.

Thymocyte egress is directly linked to TCR signaling intensity through AKT/FOXO1 inhibition of KLF2-dependent S1PR₁ expression (Hart et al., 2012), which likely prevents premature egress of autoreactive T lymphocytes (Zachariah and Cyster, 2010). Similarly, immature B lymphocyte egress is sensitive to BCR signaling intensity as it controlled CXCR4 expression. However, even when CXCR4 was overexpressed in developing B cells, a few immature B cells expressing high amounts of CXCR4 were detected in blood and spleen. Likewise, small numbers of B cells and neutrophils can be found in PB of WHIM patients, in which CXCR4 is constitutively active because of mutations that prevent its desensitization. Finally, autoreactive immature B cells can still egress from mouse and human BM, albeit inefficiently (Hartley et al., 1993; Wardemann et al., 2003). Combined, these data suggest that the BM architecture and its fenestrated vasculature generate a cell egress–permissive environment that contrasts with rather impermeable egress sites from the thymus. We suggest that the passive mode of leukocyte egress from BM evolved to facilitate the release of highly heterogeneous leukocyte subsets that include nonmotile cells, at the expense of some premature exit of developing B lymphocytes. This fundamental difference between T and B lymphocyte egress strategies from primary lymphoid organs possibly imposed an evolutionary pressure for additional extramedullary B cell developmental checkpoints (Carsetti et al., 2004).

Adult C57BL/6 (CD45.2), Boy/J (CD45.1), Itgb1^Fl/Fl, Lyz2^Cre/+, Cd19^Cre/+, and Rosa26^{FloxSTOP-ZsGreen/+} (Madisen et al., 2010) mice were from obtained from The Jackson Laboratory. MD4 and MD4.Bcl2 transgenic mice were from an internal colony (J.G. Cyster). B1-8 transgenic mice were provided by A. Haberman (Yale University, New Haven, CT). Cxcr4^Fl/− and Rosa26^{FloxSTOP-PTX/+} (Regard et al., 2007) mice were crossed with Mb1^Cre/+ mice (Pelanda et al., 2002) to generate Cxcr4^Fl/− Mb1^Cre/+ and Rosa26^{FloxSTOP-PTX/+} Mb1^cre/+ mice (all provided by J.G. Cyster). Rosa26^{FloxSTOP-PTX/+} mice were crossed to Lyz2^Cre/+ and Il7r^Cre/+ to generate Rosa26^{FloxSTOP-PTX/+} Il7r^Cre/+, Rosa26^{FloxSTOP-PTX/+} Lyz2^Cre/+, and control littermates. Il7r^Cre/+ mice were provided by H.-R. Rodewald (German Cancer Research Center, Heidelberg, Germany). For intravital imaging, Cxcr4^Fl/−, Itgb1^Fl/Fl, and Rosa26^{FloxSTOP-PTX/+} Mb1^Cre/+ were crossed with Rag1^GFP/+ mice (Kuwata et al., 1999). Mice were maintained under specific pathogen–free conditions and used according to the protocols approved by the Yale University Institutional Animal Care and Use Committee.

Mice were anesthetized with ketamine/xylazine and immobilized on a custom-built stage. Laser-scanning microscopy images were collected using a BX61WI fluorescence microscope (Olympus) and a 20× 0.95NA water immersion objective (Olympus) and dedicated single-beam TriM Scope II (LaVision Biotec) controlled by IMSpector software. The microscope was outfitted with a Chameleon Vision II Ti:Sapphire laser (Coherent) with pulse precompensation. For 4D analysis of cell migration, stacks of 13–17 optical sections with 3-µm z spacing were acquired every 20 or 30 s for 30 min with the laser tuned to a wavelength of 845–875 nm. Pharmacological antagonists AMD3100 (Tocris Bioscience), 4F-benzoyl-TN14003 (a gift from H. Tamamura, Tokyo Medical and Dental University, Tokyo, Japan), and anti–VCAM-1 (clone M/K2; Bioexpress Inc.), anti-α4 (clone PS/2; LygoCyte Pharma Inc.), and anti-CXCL12 (clone 79014; R&D Systems) blocking antibodies were injected i.v. 1–5 min before imaging. Videos in which significant tissue drifting was detected were excluded from analyses. A few videos showed small tissue drifting, which was computationally corrected with Imaris software tool “Correct Drift” (Bitplane).

The axis ratio of parenchymal GFP⁺ cells was calculated by measuring the distance between the two furthest GFP⁺ points on the x and y planes of individual cells after collapse of the z plane using Imaris. Parenchymal GFP⁺ cells that were distinguishable from neighboring cells were selected for analysis to avoid miscalculation of the axis ratio of a single GFP⁺ cell.

BM cells were flushed from femurs and tibias in DMEM (Cellgro) containing 2% FBS (Invitrogen), 5% of antibiotics (Cellgro), and Hepes (Cellgro). PB was collected from the portal vein, and erythrocytes were lysed with NH₄Cl, KHCO₃, and EDTA. Spleen was dissociated in 5 ml DMEM containing 2% FBS, 5% of antibiotics, and Hepes using a cell strainer (Thermo Fisher Scientific). Cells were counted with a Coulter Counter (Beckman Coulter). B cells were identified by staining with anti-B220 (RA3-6B2) or CD19 (1D3), anti-IgM (11/41), anti-IgD (11-26c.2a), and anti-CD93 (AA4.1) antibodies. Monocytes and granulocytes were stained with Ly6C (HK1.4), CD11b (M1/70.15), and CD115 (AFS98) antibodies, and NK cells were stained with CD3e (145-2C11) and NK1.1 (PK136) antibodies. Dead cells were excluded by staining with DAPI. Cells were analyzed by FACS (LSRII; BD).

Dye perfusion was measured by selecting two regions of interest in the BM parenchyma and FITC-conjugated labeled sinusoids using Imaris. A surface was applied to the green channel to generate voxels in regions of interest within the BM parenchyma and sinusoids over the course of the video. Perfusion of FITC-dextran into the parenchyma was assessed by measuring the number of voxels that appeared in the regions of interest over the course of a 29-min video.

Analyses of interstitial fluid flow were performed as described previously (Iliff et al., 2012; Egawa et al., 2013). In brief, flow rate within BM parenchyma was measured by injecting FITC-dextran (500 kD) or BSA–Texas red i.v. into anesthetized mice. Mice were then immediately imaged by two-photon IVM. The volumetric flow rate in the BM parenchyma was quantified by first applying a surface to the appropriate channel. Next, a three-dimensional box of a fixed volume was applied to three regions of interest. The instantaneous flow rates (on average six time points) of dye before the three-dimensional box was filled were used to calculate the mean instantaneous flow rate. The instantaneous flow rate (µm³/s) was calculated by taking the difference between volumes from consecutive time points and dividing this difference by time (60 s).

BM sinusoidal cells were labeled by injecting i.v. 0.3 µg phycoerythrin-conjugated rat anti–mouse CD19 (1D3) in 200 µl PBS. 2 min after, mice were sacrificed in a CO₂ chamber.

Mice were treated with 1 mg/ml BrdU (BD) i.v. and administered 1 mg/ml BrdU (Sigma-Aldrich) in drinking water 24 h before analysis. 1 µg PTX (List Bio Labs) was injected i.v. 24 h before analysis. BrdU-labeled cells were identified staining with anti-BrdU FITC according to the manufacturer’s protocol. B cells were identified by staining with anti-B220 (RA3-6B2) or CD19 (ebio1D3), anti-IgM (11/41), anti-IgD (11-26.c.2a), and anti-CD93 (AA4.1). Monocytes, NK cells, and granulocytes were identified by staining for anti-Ly6c (HK1.4), anti-CD11b (M1/70), anti-CD3ε (145-2C11), anti-NK1.1 (PK136), and anti-CD115 (AFS98). After staining, cells were analyzed by FACS (LSRII).

Approximately 1.5 × 10⁶ total BM cells from Ly5.1⁺ donors were mixed with 1.5 × 10⁶ total BM cells from adult Boy/J (Ly5.2⁺) mice and were transferred into adult Boy/J mice that had been exposed to two rounds of 6.35 Gy separated by 3 h. Chimeras were analyzed at least 6 wk after reconstitution.

Chemotaxis assays were performed using 10⁶ BM or spleen cells incubated for 30 min with 1× DMEM containing 0.5% fatty acid–free BSA (EMD Biosciences), 5% of antibiotics, l-glutamine (Cellgro), and Hepes. Cells were then allowed to migrate through 5-µm-pore–sized transwells (Corning) toward soluble CXCL12 (R&D Systems), 2-AG (Cayman), or CXCL13 (PeproTech) for 3 h at 37°C. Cells were collected, stained, and resuspended in 40 ml of staining buffer and analyzed by flow cytometry for 40 s.

The R334X mutation in CXCR4 cytoplasmic tail was cloned into mammalian retroviral vector (pMSCV) upstream of an IRES-truncated GFP cassette as a reporter. Sanger DNA sequencing reaction verified truncated CXCR4 sequences. Phoenix 293T cells were transfected with MSCV retroviral constructs with Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. For mutant Cxcr4 (R33x) transduction of hematopoietic stem cells, BM cells were harvested 5 d after i.p. injection of 5-fluoracil (Sigma-Aldrich) and cultured with IL-3, IL-6, and 100 ng/ml mSCF (PeproTech). BM cells were then spin-infected twice with a retroviral constructs and injected i.v. into lethally irradiated Boy/J recipients. Mice were analyzed at least 5 wk after reconstitution.

10⁶ BM or spleen cells were treated with saline (0.9% NaCl) or AMD3100 for 1 h at 37°C. Cells were then removed from incubation and used in transwell migration assays.

BM cells isolated from femurs and tibias were prepared at 10⁷ cells/ml in DMEM containing 10% FBS, 10 mM Hepes, and a cocktail of penicillin and streptomycin (50 U/liter and 50 µg/liter, respectively). Approximately 2 × 10⁶ cells were left unstimulated or stimulated with 20 µg/ml F(ab′)2 fragment goat anti–mouse IgM (Jackson ImmunoResearch Laboratories, Inc.), with 20 µg/ml HEL (Sigma-Aldrich), or with 1 mg/ml NP-BSA (Biosearch Technologies, Inc.), at 37°C and 5% CO₂ for the indicated time points in the Fig. 6 legend.

MD4 and MD4.Bcl2 transgenic mice were stimulated with 10 mg HEL for 6 h. Mice were injected with 0.3 µg phycoerythrin-conjugated rat anti–mouse CD19 (1D3) in 200 µl PBS before sacrifice. Cells isolated from femur and tibia BM, PB (portal vein), and spleen were counted in a Coulter Counter and stained to detect B cell subsets, as previously described in Tissue preparation, cell enumeration, and antibodies.

Student’s two-tailed t test or χ² test was performed using GraphPad Software Prism 6 or Microsoft Excel. A value <0.05 was considered significant.

Video 1 shows intravital imaging of developing B cell migration before and after treatment with AMD3100 (25 min). Video 2 shows intravital imaging of Cxcr4-deficient developing B cell migration (30 min). Video 3 shows intravital imaging of Cxcr4-deficient developing B cell migration before and after treatment with AMD3100 (30 min). Video 4 shows intravital imaging of developing B cell migration before and after treatment with anti–VCAM-1 (30 min). Video 5 shows intravital imaging of Itgb1-deficient developing B cell migration (30 min). Video 6 shows intravital imaging of developing B cell migration before and after treatment with anti–VCAM-1 (30 min). Video 7 shows intravital imaging of FITC-conjugated dextran perfusion into the BM parenchyma (29 min). Video 8 shows intravital imaging of PTX-expressing B lineage cell migration in BM (30 min). Video 9 shows intravital imaging of motile PTX-expressing B lineage cell migration in BM (30 min). Video 10 shows intravital imaging of WT and PTX-expressing B lineage cell egress from BM.

We thank Dr. Heins-Reimer Rodewald for sharing Il7ra^Cre/+ mice. We thank Flavia O. Pinho (Yale University, New Haven, CT) for technical assistance with retroviral DNA constructs, Dr. Hirokazu Tamamura for 4F-benzoyl-TN14003, and David Gonzalez and Ann Haberman (Yale University) for assistance with the two-photon microscopy. We are grateful to S.R. Schwab (New York University, New York, NY) and M. Carmona (Yale University) for critical reading of the manuscript.

T.C. Beck is the recipient of a predoctoral National Science Foundation fellowship; A.C. Gomes is supported by a predoctoral fellowship from the Foundation for Science and Technology, Ministry of Education, Portugal. This work was partially supported by the National Institutes of Health (grants R56AI098996 and RO1AI113040).

The authors declare no competing financial interests.