Bone Marrow Endothelial Cells Take Up Blood-Borne Immune Complexes via Fcγ Receptor IIb2 in an Erythropoietin-Dependent Manner Free

Antibodies bind to their specific Ags and form immune complexes (ICs) to neutralize infectious organisms and harmful molecules. ICs also induce immune reactions including complement activation and opsonization, leading to the clearance of target Ags. However, excess and/or the incomplete removal of ICs may induce vasculitis and tissue inflammation with the deposition of ICs (1, 2). This effect is the representative pathology of type III hypersensitivity and is observed in patients with autoimmune diseases (such as systemic lupus erythematosus and acute poststreptococcal glomerulonephritis) who develop large amounts of ICs constituting Abs against autoantigens and infectious microbes, respectively (2–4). Thus, the prompt and efficient clearance of ICs from the blood circulation is important for the elimination of harmful Ags as well as for the prevention of systemic tissue inflammation.

ICs are captured and digested mainly in the liver by the reticuloendothelial system, which consists of Kupffer cells and sinusoidal endothelial cells (ECs) (5–8). The endocytosis of ICs by these cells is mediated by FcγRs, which are receptors for the Fc portion of IgG (9). The four different classes of FcγRs include three activating FcγRs known as FcγRI, FcγRIII, and FcγRIV and one inhibitory FcγR named FcγRIIb in mice, and each FcγR is expressed on various cell types, including immune cells at various combinations (10). Among the cells, Kupffer cells, which are phagocytic cells in the liver, express both activating and inhibitory FcγRs, whereas liver sinusoidal ECs express only FcγRIIb, and both cell types efficiently endocytose ICs but not monomeric IgGs in a clathrin-dependent manner (5, 11–13). Additionally, the decay of i.v.-injected ICs in the blood is delayed in Fcgr2b-deficient mice compared with control mice, suggesting a role for FcγRIIb in the clearance of ICs (13).

The bone marrow (BM) is a hematopoietic organ that constitutively produces multilineage hematopoietic cells. BM endothelial cells (BMECs) also constitute sinusoidal vasculature and have unique function in constitutive hematopoiesis as a part of the hematopoietic niche by producing various cytokines/chemokines and adhesion molecules (14, 15). Thus, BMECs are functionally distinct from ECs in other organs; however, it has not been addressed whether BMECs have a capacity to clear ICs like liver sinusoidal ECs. Recently, it was reported that the erythropoietin receptor (Epor) is expressed in BMECs (16, 17), and we demonstrated that BMECs induce immature and mature B cell egress from the BM by decreasing Cxcl12 and Vcam1 expression upon erythropoietin (Epo) stimulation (16). Epo is the major hematopoietic hormone for erythropoiesis and is produced by the kidneys in response to hypoxia (18, 19). In addition to its erythropoietic and antihypoxic functions, Epo has immunomodulatory functions, including the inhibition of proinflammatory cytokine production and the promotion of apoptotic cell clearance in macrophages (20–22).

In the current study, we found that sinusoidal BMECs express FcγRIIb and that Epo administration induces the upregulation of FcγRIIb expression in BMECs. The administration of Epo causes a remarkable increase in the capacity of capturing ICs in BMECs via FcγRIIb and promotes the rapid removal of ICs in the blood circulation, suggesting a novel role of BMECs in enhancing the clearance of ICs under anemic conditions.

C57BL/6 mice (Japan SLC), W/W^v mice, and their wild-type (WT) control (WBB6F1/Kit-Kit^W/Kit^W-v/Slc and WBB6F1/Kit-Kit⁺/Kit⁺/Slc; Japan SLC) (23) and FcγRIIb knockout (KO) mice (Oriental Bioservice) (24) were maintained in standard pathogen-free conditions. Epor^+/−::HG1-Epor transgenic (Tg) mice (25) were provided by the RIKEN BioResource Research Center through the National BioResource Project, Japanese Ministry of Education, Culture, Sports, Science and Technology, Japan. Eight-week-old mice were analyzed unless specified otherwise. The induction of anemia and treatment with Epo were previously described (16). In brief, to induce hemorrhagic anemia, mice were bled from the retro-orbital sinus to remove 400 μl of blood under isoflurane anesthesia and i.p. injected with 500 μl of PBS for volume replacement. Bleedings were performed for three consecutive days and analyzed on day 4. To induce hemolytic anemia, mice were i.p. injected with 50 mg/kg phenylhydrazine (PHZ; Sigma-Aldrich) for two consecutive days and analyzed on day 4. For Epo treatment, mice were i.p. injected with either 3000 U/kg recombinant human Epo (Epoetin β; Chugai) or PBS (control group) for three consecutive days and analyzed on day 4. All experiments were performed according to the guidelines of the Animal Research Committee, Graduate School of Medicine, Kyoto University.

BM cells were flushed out from the femoral and tibial bones and suspended in media. To obtain ECs and mesenchymal stromal cells (MSCs), BM cells were incubated in DMEM high glucose with 2% FCS, 5 mg/ml collagenase type I (Life Technologies), and 100 μg/ml DNase I (Worthington Biochemical) at 37°C for 20 min before cell suspension, as previously described (16). To obtain ECs from liver, spleen, lung, kidney, and thymus, these organs were dissected and incubated for 30 min before cell suspension in the media. To obtain osteoblasts, marrow-depleted bones were dissected and crushed and then incubated for 45 min before cell suspension in the media. Hematocrit values in peripheral blood were measured using a Celltac α Analyzer (Nihon Kohden).

After lysing erythrocytes with ammonium-chloride-potassium (ACK) buffer, BM and spleen cells were suspended in PBS with 2% FCS and 5 mM EDTA. After Fc blocking, the samples were stained with Abs. Cells were analyzed using a FACS Canto II cytometer (BD Biosciences) or SA3800 spectral analyzer (Sony), and the data were analyzed with Flow Jo software (Flow Jo). To evaluate the expression of FcγRIIb, Fc blocking was skipped. ECs were defined as CD45⁻Ter119⁻CD31⁺Sca1⁺, as previously described (16). Liver Kupffer cells were defined as CD45⁺F4/80⁺.

To sort BMECs and MSCs, collagenase-digested BM cells were Fc blocked and stained with PE-conjugated anti-CD45 and anti-Ter119 Abs and biotin-conjugated anti-PDGFRβ Abs. After washing, the cells were incubated with anti-PE AutoMACS beads (Miltenyi Biotech), and then the majority of hematopoietic cells were depleted using an AutoMACS Pro Separator (Miltenyi Biotech). Negatively collected cells were stained with FITC-conjugated anti-CD31 and PECy7-conjugated anti-Sca1 Abs and with allophycocyanin-conjugated streptavidin. Cell sorting was performed using a FACS Aria system (BD biosciences). To sort the hematopoietic cells from the BM and spleen, the AutoMACS depletion step was skipped. Neutrophil, monocytes, and B cells were sorted from the BM, and NK cells were prepared from the spleen.

The following fluorescent or biotinylated Abs were used: anti-B220 (clone RA3-6B2; BioLegend), anti-CD3e (clone 145-2C11; eBioscience), anti-CD11b (clone M1/70; Tonbo Biosciences), anti-CD16 (FcγRIII, clone 275003; R and D), anti-CD16/32 (FcγRIIb and FcγRIII, clone 2.4G2; BD Pharmingen), anti-CD31 (clone 390; BioLegend), anti-CD45 (clone 30-F11; BioLegend), anti-CD49b (clone DX5; eBioscience), anti-CD51 (clone RMV-7; BioLegend), anti-CD71 (clone R17217; eBioscience), anti-CD115 (clone AFS98; BioLegend), anti-F4/80 (clone BM8; BioLegend), anti–Gr-1 (clone RB6-8C5; BioLegend), anti-NK1.1 (clone PK136; Tonbo Biosciences), anti-PDGFRβ (goat polyclonal; R and D), anti-Sca1 (clone D7; BioLegend), anti-Ter119 (clone TER119; BioLegend), and anti-VEGFR3 (goat polyclonal; R and D).

Femoral bones were fixed in 4% paraformaldehyde for 2 h at room temperature and equilibrated in 30% sucrose/PBS overnight. Fixed samples were embedded in super cryoembedding medium (Section laboratory) and frozen in cold hexane. Sections were generated using the Kawamoto method (26). Cryostat sections (6-μm thick) were first blocked with 5% normal donkey serum/PBS and then stained with the following primary and secondary Abs: anti-CD105 (clone MJ7/18; BioLegend), anti–CD32-B (FcγRIIb, goat polyclonal; Santa Cruz), Alexa Fluor 488 (Alexa488)–conjugated anti-rat IgG (donkey polyclonal; Invitrogen), and Alexa Fluor 555–conjugated anti-goat IgG (donkey polyclonal; Invitrogen). Sections were mounted with Mowiol (Sigma-Aldrich), and microscopic observation was performed using an Axiovert 200 M inverted microscope (Zeiss) or BZ-X810 (Keyence).

Sorted cells were purified with an RNeasy Micro Kit (Qiagen), and cDNA was synthesized using SuperScript III (Thermo Fisher Scientific) and Oligo-dT primers (Thermo Fisher Scientific) according to the manufacturer’s instructions. PCR was performed with Ex Taq (Takara) or KOD FX (Toyobo). Primer sequences are as follows: Actb, 5′-TCCTGTGGCATCCATGAAACT-3′ and 5′-CGCAGCTCAGTAACAGTCCGCC-3′ and Fcgr2b, 5′-CCAAGCCTGTCACCATCACT-3′ and 5′-TGGCTTGCTTTTCCCAATGC-3′.

cDNA was quantified by real-time PCR using a Light Cycler 480 II (Roche) and SYBR Green Master Mix (Roche). The expression level of each gene was normalized to that of Actb in each sample.

Purified polyclonal rabbit anti-BSA IgG (Invitrogen) was conjugated with Alexa488 using an Alexa488 protein labeling kit (Invitrogen), following the manufacturer’s instructions, and dialyzed with PBS by using a Slide-A-Lyzer dialysis cassette (Thermo Fisher Scientific). ICs were prepared by incubating in BSA (A2058; Sigma-Aldrich) with Alexa488-conjugated anti-BSA Abs at a molar ratio of five to six Abs to one Ag for 30 min at 37°C. Control IgGs were prepared by the same procedure without BSA. WT or FcγRIIb KO mice pretreated with Epo or PBS were injected with 100 μg of Alexa488-conjugated ICs or IgG and analyzed 30 min after injection. To analyze the retention of incorporated ICs or IgGs, the mice were i.v. injected with Alexa488-labeled ICs or IgGs and analyzed 24 h, 1 wk, and 2 wk after injection. Epo was administered every other day after the IC or IgG injection. To evaluate the phagocytotic capacity of large particles, mice were i.v. injected with 20 μl of 1-μm latex beads (fluorescent yellow-green; Sigma-Aldrich) and analyzed 30 min after injection.

Collagenase-digested BM cells were stained with PE-conjugated anti-CD45 and anti-Ter119, allophycocyanin-conjugated anti-CD31, and allophycocyanin-Cy7–conjugated anti-Sca1 Abs and sorted using an AutoMACS Pro Separator and a FACS Aria system as described above. The sorted BMECs were cultured on a multiwell glass-bottom dish (Matsunami) with the EC growth medium MV 2 Kit (PromoCell). The dishes were precoated with fibronectin from human plasma (Wako) for 2 h at 37°C. Epo (10 U/ml) or PBS was added on day 2 and coincubated with ICs (BSA/rabbit anti-BSA IgG, 22 μg/ml) for 15 min or 2 h at 37°C on day 3. The cells were fixed in 4% paraformaldehyde for 15 min at room temperature, permeabilized in 0.2% Triton X for 15 min, and blocked with 5% normal donkey serum/PBS. They were stained with anti–LAMP-1 Ab (clone 1D4B; BioLegend) followed by Alexa488-conjugated anti-rabbit IgG (donkey polyclonal; Invitrogen) and Cy3-conjugated anti-rat IgG (donkey polyclonal; Jackson ImmunoResearch) Abs and DAPI. Microscopic observation was performed by using an FV1000 confocal microscope (Olympus).

Preformed ICs of BSA/rabbit anti-BSA IgG (100 μg) were i.v. injected into Epo- or PBS-treated mice, and the blood was consecutively collected from the retro-orbital sinus 0, 10, 30, and 60 min after the injection. The plasma concentration of the ICs was measured by ELISA. The ICs in the plasma were captured by mouse anti-BSA mAbs (Abcam) and detected with HRP-conjugated mouse anti-rabbit IgG L chain–specific mAbs (Jackson Immunoresearch). Normal mice serum (Fujifilm) was used to prevent nonspecific signals. KPL SureBlue (Seracare) was used for the enzyme reaction, which was stopped by TMB Stop Solution (Seracare). Signals were read by EnVision 2104 (PerkinElmer).

The significance of the results was assessed using two-tailed Student t tests for the comparison of two groups and one-way ANOVA followed by Tukey test for comparisons of more than two groups and calculated with GraphPad Prism (GraphPad Software). A p value < 0.05 was considered significant. All data are presented as means and SEM.

Flow cytometric analysis indicated that ∼60% of the CD31⁺ ECs in adult BM were stained with 2.4G2 Ab, which recognizes both FcγRIIb (CD32) and FcγRIII (CD16) (27), with two peaks, 2.4G2^low and 2.4G2^high (Fig. 1A). Both the 2.4G2^low and 2.4G2^high signals were completely abolished in Fcgr2b-deficient mice, indicating that the BMECs specifically expressed FcγRIIb rather than FcγRIII (Fig. 1A). We confirmed that essentially all liver ECs and ∼40% of splenic ECs were also positive for 2.4G2 staining, whereas ECs in other tissues, including lung, kidney, and thymus, were hardly stained with 2.4G2 (Supplemental Fig. 1A). In the BM, other stromal cells, including CD51⁺ osteoblasts and PDGFRβ⁺ MSCs, were not stained with 2.4G2 at all (Fig. 1B). Immunohistochemical staining indicated that the signal with anti-FcγRIIb Ab was detected in sinusoid CD105⁺ ECs in the BM, which was completely abolished in Fcgr2b-deficient mice (Fig. 1C). FcγRIIb staining was not detected in the large central veins or αSMA⁺ arteries, indicating that the expression of FcγRIIb was specific to sinusoidal BMECs (Supplemental Fig. 1B). Fcgr2b has two splice variants, Fcgr2b1 and Fcgr2b2, the latter lacking the first cytoplasmic exon (28, 29). We confirmed that various types of immune cells express FcγR families with different combinations: monocytes expressed all types of Fcgrs, neutrophils predominantly expressed Fcgr3 and Fcgr4, B cells expressed Fcgr2b1, and NK cells expressed Fcgr2b1, Fcγr3, and Fcγr4 (Supplemental Fig. 1C) (3, 29, 30). We further found that BMECs almost exclusively expressed Fcgr2b2 among FcγR family genes (Fig. 1D, Supplemental Fig. 1C). These results indicate that sinusoidal BMECs specifically expressed FcγRIIb2.

Next, we examined the FcγRIIb2 expression in BMECs during development. In newborn and 2-wk-old mice, FcγRIIb2 expression in BMECs was negligible but began to be detected in 4-wk-old mice and increased thereafter with age (Fig. 2A); in mice that were 39 wk of age and older, most BMECs expressed FcγRIIb2 (Fig. 2A). Immunohistochemical staining revealed that FcγRIIb2 expression increased as the sinusoidal structures were established with age in the BM (Fig. 2B). This pattern contrasts that of the liver, in which FcγRIIb2 expression was seen in the great majority of ECs in newborns, and >90% of liver ECs exhibited FcγRIIb2 expression in 2-wk-old mice (Supplemental Fig. 2). These results suggest that FcγRIIb2 expression in BMECs appears to be associated with the development and maturation of the sinusoidal vasculatures in the BM after birth.

We have reported that BMECs express Epor and that Epo alters the expressions of various cytokines and adhesion molecules in BMECs (16). Therefore, we investigated the possible effects of Epo on FcγRIIb2 expression in vivo. First, we examined the effects of acute hemorrhage or PHZ administration in adult mice, which caused severe acute anemia (Supplemental Fig. 3A) followed by an increase in serum Epo concentration (16, 31). Both anemia-inducing treatments resulted in a significant increase in FcγRIIb2 expression in BMECs (Fig. 3A, Supplemental Fig. 3B). Additionally, a more-obvious upregulation of FcγRIIb2 expression was observed in Kit mutant W/W^v mice, which exhibit chronic anemia (Fig. 3A, Supplemental Fig. 3B). To see whether the effects were attributed to the Epo induction in vivo, we directly injected Epo into healthy mice and evaluated the FcγRIIb2 expression in BMECs. As expected, the FcγRIIb2 expression in BMECs was remarkably increased in both proportion and intensity and in mRNA (Fig. 3B, Supplemental Fig. 3C, 3D). In contrast, the patterns of FcγRIIb expression in the ECs of other organs, including the liver, spleen, lung, and kidney, and in Kupffer cells were not affected at all by the Epo administration (Supplemental Fig. 3E). FcγRIII expression detected by FcγRIII-specific Ab was minimally affected both in BMECs and Kupffer cells by Epo administration (Supplemental Fig. 3F). To address whether the Epo-induced FcγRIIb upregulation depended on the EpoR expression on BMECs, we employed Epor^−/−::HG1-Epor Tg mice, in which the GATA1 promotor-driven Epor transgene was expressed in Epor-deficient mice (25). Whereas control (Epor^+/−::HG1-Epor Tg) mice expressed endogenous Epor in erythroblasts and, to a lesser extent, BMECs, only erythroblasts exhibited the Epor transgene in Epor^−/−::HG1-Epor Tg mice (Supplemental Fig. 3G). Upon Epo administration, a significant increase of FcγRIIb2 expression in BMECs was observed in control mice but not in Epor^−/−::HG1-Epor Tg mice (Fig. 3C). Furthermore, FcγRIIb2 expression in the PHZ-induced anemic condition was significantly lower in Epor^−/−::HG1-Epor Tg mice than in control mice (Supplemental Fig. 3H). These results strongly suggest that Epo induced the increased cell surface expression of FcγRIIb2 in BMECs directly via EpoR on the BMECs in vivo.

Because liver sinusoidal ECs expressing FcγRIIb2 are capable of the constitutive uptake of ICs (12, 13), we next addressed whether BMECs had the capacity of incorporating blood-borne soluble ICs in vivo. To this end, we i.v. injected soluble Alexa488-labeled IgG or preformed ICs of BSA/Alexa488-labeled anti-BSA IgG into adult mice pretreated with Epo or PBS as a control and 30 min later analyzed the Alexa488 signals associated with the ECs from various tissues with flow cytometry (Supplemental Fig. 4A). In control mice, liver ECs showed a high Alexa488 signal following IC injection, but not soluble IgG injection, as previously reported (5, 12), whereas BMECs and splenic ECs showed little or no Alexa488 signal despite their expression of FcγRIIb2 (Fig. 4A). In contrast, ∼20% of BMECs exhibited a clear Alexa488 signal following the soluble IC injection in Epo-treated mice, although the Alexa488 signals remained negative in splenic ECs and unaffected in liver ECs and Kupffer cells (Fig. 4A). Importantly, the Alexa488⁺ BMECs were confined to the FcγRIIb2^high EC fraction of IC-injected mice (Fig. 4B), suggesting that FcγRIIb2^high cells take up ICs. In contrast to soluble ICs, 1-μm latex beads were incorporated in Kupffer cells much more efficiently than ECs in liver and the BM (Supplemental Fig. 4B). We also observed that the Alexa488 signal of ICs was sustained stably for at least 2 wk after IC injection (Supplemental Fig. 4C).

We then examined whether Epo directly affects BMECs using an in vitro culture system. To this end, sorted BMECs were cultured on fibronectin-coated dishes in EC growth medium and fed Epo (10 U/ml) or PBS on day 2, followed by the addition of ICs (22 μg/ml) on day 3 for 15 min or 2 h. We observed Alexa488⁺ aggregates of ICs in the cytoplasm (Supplemental Fig. 4D). The percentage of IC-incorporated BMECs was remarkably increased by the Epo treatment after 2 h of coincubation with ICs (Fig. 4C). These results suggest that Epo directly enhanced the capacity of incorporating blood-borne ICs in BMECs.

Next, we addressed whether the incorporation of soluble ICs in BMECs is mediated by FcγRIIb2 by using Fcgr2b-deficient mice. Although the Epo treatment of WT mice caused a remarkable increase in the uptake of i.v.-injected Alexa488-labeled ICs in BMECs compared with control WT mice, Fcgr2b-deficient mice exhibited negligible IC uptake irrespective of the Epo treatment (Fig. 5A, 5B). Furthermore, whereas liver ECs in control WT mice showed substantial IC uptake, which was minimally affected by Epo treatment, the activity was almost completely abrogated in Fcgr2b-deficient mice (Fig. 5A). These results confirmed that Epo caused a selective enhancement of soluble IC uptake in BMECs via FcγRIIb2.

Finally, we questioned whether the Epo-induced incorporation of ICs in BMECs promotes the elimination of ICs from the bloodstream. To this end, we i.v. injected 100 μg of preformed ICs of BSA/anti-BSA IgG into WT mice pretreated with Epo or PBS and measured the plasma concentration of ICs 0, 10, 30, and 60 min later. The plasma concentration of ICs in both groups was comparable immediately after the IC injection (0 min), but it was significantly lower in the Epo-treated mice than in control mice 10 min after the IC injection (Fig. 6). The plasma concentration of ICs in both groups decreased to the same levels 60 min after the IC injection (Fig. 6). Collectively, these results strongly suggest that Epo promotes the rapid clearance of soluble ICs in the bloodstream by inducing the IC incorporation capacity selectively in BMECs.

In this study, we found that sinusoidal BMECs express FcγRIIb2 and that the expressions are upregulated by Epo. BMECs exhibit an Epo-induced incorporation of soluble ICs via FcγRIIb2, and Epo itself promotes the removal of ICs from the blood circulation. Hence, our study suggests that BMECs contribute to the clearance of ICs from the blood when the serum Epo level is increased in anemic conditions.

The vascular system in the BM plays important roles in hematopoietic homeostasis, including postnatal definitive hematopoiesis in vascular niches and trafficking of hematopoietic cells (14, 15). In the current study, we found that ∼60% of BMECs in healthy adult mice exhibited FcγRIIb2 expression, which is confined to the ECs constituting sinusoidal vasculatures. The ECs of sinusoidal vasculatures in the liver and spleen also expressed FcγRIIb2, whereas ECs in other tissues without the sinusoidal system rarely did, indicating the expression of FcγRIIb2 is one of the unique characteristics of ECs constituting the sinusoidal vascular system. Of note, unlike in liver, in which the great majority of ECs already exhibit FcγRIIb2 in neonates, FcγRIIb2⁺ ECs in the BM began to develop a few weeks after birth and steadily increased with age, apparently correlating with the establishment of the sinusoidal vasculatures. It remains to be seen whether the development of FcγRIIb2⁺ BMECs represent trans-differentiation from regular ECs or neovascularization under the developing hematopoietic microenvironment.

Most notably, we found that both the proportion of FcγRIIb2⁺ ECs and intensity of FcγRIIb2 expression were significantly increased in anemic conditions, be it acute anemia due to bleeding or hemolysis or chronic anemia due to Kit insufficiency (W/W^v). We previously reported that BMECs uniquely bore functional EpoR, and our current results indicated that the administration of Epo increased the EC population with higher FcγRIIb2 expression. Furthermore, these effects of Epo and anemia on BMECs were almost completely abrogated in Epor^−/−::HG1-Epor Tg mice, which do not express EpoR in BMECs. Hence, our findings suggest the direct effects of Epo on BMECs, with Epo interacting with the EpoR expressed on BMECs, although an indirect effect of Epo and/or pathways other than EpoR signaling could be also involved. Epo showed no significant effects on the FcγRIIb2 expression in liver and spleen ECs, probably because of the lack of EpoR expression (16). Our previous results indicated that Epo stimulation induced STAT5 phosphorylation and changes in the gene expressions of various cytokines and adhesion molecules, such as Cxcl12 and Vcam1 in BMECs (16). We also found that the upregulation of Fcgr2b is downstream of the EpoR signaling. Although Epor^−/−::HG1-Epor Tg mice showed no increase in the proportion of FcγRIIb2⁺ ECs in the BM after Epo administration, the postnatal development of FcγRIIb2⁺ BMECs per se was comparable with that in control mice. Thus, Epo does not seem to be crucially involved in the physiological development of these cells.

There are two functional splice variants in FcγRIIb: FcγRIIb1 is expressed on B cells and suppresses B cell activation via an ITIM motif on binding ICs, and FcγRIIb2 is expressed on dendritic cells and monocytes to promote the incorporation of bound ICs (28, 29). The current results indicated that BMECs selectively express FcγRIIb2. Unlike liver ECs, BMECs in healthy adult mice showed a minimal uptake of blood-borne soluble ICs in vivo, whereas BMECs in Epo-treated mice showed a significant uptake of ICs. Considering that the signal intensity of the incorporated ICs was correlated with the expression levels of FcγRIIb2, the difference in the capacity of IC incorporation in the BMECs and liver ECs may be due to the lower expression levels of FcγRIIb2 in the BMECs than liver ECs in steady-state. Other mechanisms, such as endocytosis activated by FcγRIIb2 and/or EpoR-mediated signaling, could be also involved.

The current results further indicate that the soluble ICs were incorporated into the cytosols of BMECs and were sustained stably for >2 wk. It was reported that FcγRIIb-bound ICs were directed to a nondegradative intracellular vesicular compartment and thus sustained for long periods (12, 32). Further, a portion of these intact ICs were continuously recycled back to the cell surface to be accessible to naive B cells (33). It remains to be investigated whether the BMECs that incorporated soluble ICs via FcγRIIb2 and stably retained them intracellularly can interact with naive B cells via the ICs recycled to the cell surface. In this context, we reported that BMECs stimulated by Epo promote the release of naive B cells from the BM to the periphery via the downregulation of Cxcl12 and Vcam1 expression (16). As such, it may be an intriguing possibility that sinusoidal BMECs incorporate soluble ICs, particularly those involving T-independent Ags, to make the ICs accessible to developing naive B cells and further promote the peripheralization of such primed B cells from the BM in anemic conditions.

In addition to its erythropoietic activity, Epo shows diverse immunomodulatory effects by directly binding to EpoR expressed on certain immune cells, particularly macrophage-lineage cells (34, 35). Thus, for instance, Epo promotes the apoptotic cell clearance by macrophages that ameliorate lupus disease progression and also suppresses inflammatory tissue reactions by inhibiting NF-κB signaling in macrophages (20, 21). In the current study, we found that Epo significantly enhanced the expression of FcγRIIb2 selectively in sinusoidal BMECs and strongly induced the capacity of incorporating soluble blood-borne ICs by these cells in vivo. Furthermore, we demonstrated that Epo administration induces the rapid clearance of ICs in blood circulation. Overall, our results suggest that BMECs play an important role in eliminating excess harmful blood-borne ICs, in particular under anemic conditions. We speculate that bleeding is sometimes accompanied by infections, and the increased IC incorporation in BMECs may be one of the immunomodulatory functions of Epo (20–22). This mechanism could also contribute to removing harmful ICs in autoimmune hemolytic anemia (36).

In summary, we demonstrated that BMECs exhibit a capacity for FcγRIIb2-mediated IC endocytosis by Epo stimulation, thereby promoting the clearance of ICs in the circulation. These findings suggest novel functions of BMECs in the clearance of blood-borne ICs as well as of Epo on immunomodulation and cross-talk with the immune system.

We are grateful to Dr. T. Takai for helpful comments, C. Inoue for performing experiments, other members of our laboratory for technical assistance, and Dr. P. Karagiannis for proofreading.

The authors have no financial conflicts of interest.