CD11b+ Gr1+ bone marrow cells ameliorate liver fibrosis by producing interleukin-10 in mice


  • Potential conflict of interest: Nothing to report.

  • Supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (2011-0029328) and grants of the Korean Health Technology R&D Project, Ministry of Health & Welfare, Republic of Korea (A111345 and A111498) and NIH Grant RO1DK56621.


Clinical trials and animal models suggest that infusion of bone marrow cells (BMCs) is effective therapy for liver fibrosis, but the underlying mechanisms are obscure, especially those associated with early effects of BMCs. Here, we analyzed the early impact of BMC infusion and identified the subsets of BMCs showing antifibrotic effects in mice with carbon tetrachloride–induced liver fibrosis. An interaction between BMCs and activated hepatic stellate cells (HSCs) was investigated using an in vitro coculturing system. Within 24 hours, infused BMCs were in close contact with activated HSCs, which was associated with reduced liver fibrosis, enhanced hepatic expression of interleukin (IL)-10, and expanded regulatory T cells but decreased macrophage infiltration in the liver at 24 hours after BMC infusion. In contrast, IL-10–deficient (IL-10−/−) BMCs failed to reproduce these effects in fibrotic livers. Intriguingly, in isolated cells, CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+ BMCs expressed more IL-10 after coculturing with activated HSCs, leading to suppressed expression of collagen and α-smooth muscle actin in HSCs. Moreover, these effects were either enhanced or abrogated, respectively, when BMCs were cocultured with IL-6−/− and retinaldehyde dehydrogenase 1−/− HSCs. Similar to murine data, human BMCs expressed more IL-10 after coculturing with human HSC lines (LX-2 or hTERT), and serum IL-10 levels were significantly elevated in patients with liver cirrhosis after autologous BMC infusion. Conclusion: Activated HSCs increase IL-10 expression in BMCs (CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+ cells), which in turn ameliorates liver fibrosis. Our findings could enhance the design of BMC therapy for liver fibrosis. (HEPATOLOGY 2012;56:1902–1912)

For the past decade, clinical trials and experimental studies have suggested that infusion therapy of whole bone marrow cells (BMCs) has beneficial effects toward liver regeneration, injury, and fibrosis/cirrhosis by stimulating the proliferation of hepatocytes, increasing progenitor cells, and enhancing matrix degradation.1-3 However, the underlying mechanisms are unknown, in part because whole BMCs contain a wide range of cell types, including several types of stem and precursor cells of monocytic and granulocytic lineages.4 Events associated with hepatic fibrosis are well characterized, notably the excessive production of extracellular matrix (ECM) by activated hepatic stellate cells (HSCs).5 Activated HSCs produce not only huge amounts of ECMs including collagen, but also other fibrosis-related mediators including transforming growth factor (TGF)-β1, interleukin (IL)-6, IL-10, and retinoic acid.5-7 Interestingly, treatments with TGF-β1, IL-6, and retinoic acid can differentiate naïve T cells into regulatory T cells (Tregs) or Th-17 cells in vitro, in which TGF-β1 is considered as an initial driver of this commitment.8 Moreover, activated HSCs produce these mediators implicating activated HSCs in immune regulation.

Recent studies underscore the immunoregulatory potential of HSCs, wherein they can act as intrahepatic antigen-presenting cells to activate T cells, natural killer (NK) cells, and NK T cells9, 10 and are also involved in the induction of CD11b+Gr1+ myeloid-derived suppressor cells (MDSCs) and CD4+CD25+Foxp3+ Tregs in an interferon-γ and retinoic acid–dependent manner, respectively.11, 12 MDSCs expressing both markers of CD11b and Gr1 are now appreciated as a negative regulator of immune responses in cancer and other diseases. In addition, MDSCs are closely related to the induction of Tregs in the tumor microenvironment, which could produce IL-10 through the activity of the transcription factor, Foxp3.13-15 Moreover, IL-10 is recognized as an anti-inflammatory and antifibrotic mediator.5, 6 These findings provide a rationale for the possible immunoregulatory role of HSCs in vivo during BMC infusion therapy. In fact, infused BMCs have been detected in fibrotic areas within 24 hours and can replace 25% of recipient hepatocytes by 4 weeks.16 However, the mechanisms underlying the effects of BMCs are still uncertain, and most studies of BMC infusion therapy have focused on hepatocyte regeneration and ECM degradation as long-term effects of BMCs (at least 2 weeks after BMC infusion) in liver fibrosis.1, 2 Contrary to these previous findings, in the current study, we show that HSCs directly interact with infused BMCs, especially CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+ cells among whole BMC isolates at an early phase in vivo (i.e., within 24 hours). This interaction drives production of IL-10 in both types of cells, leading to increased Tregs in the recipient liver, which attenuates fibrosis.


α-SMA, α-smooth muscle actin; BMC, bone marrow cell; CCl4, carbon tetrachloride; COL1A1, type 1 collagen alpha 1; ECM, extracellular matrix; FACS, fluorescence-activated cell sorting; GFP, green fluorescence protein; HSC, hepatic stellate cell; IL, interleukin; MCP-1, monocyte chemoattractant protein-1; MDSC, myeloid-derived suppressor cell; MMP, matrix metalloproteinase; MNC, mononuclear cell; mRNA, messenger RNA; NK cell, natural killer cell; qRT-PCR, quantitative reverse-transcription polymerase chain reaction; RALDH1, retinaldehyde dehydrogenase 1; TGF, transforming growth factor; Tregs, regulatory T cells; WT, wild-type.

Materials and Methods


Male C57BL/6, IL-6−/−, IL-10−/−, and green fluorescence protein (GFP)-transgenic mice were purchased from The Jackson Laboratory (Bar Harbor, ME). B6/SJL (CD45.1) mice were purchased from Taconic (Germantown, NY). Retinaldehyde dehydrogenase 1 (RALDH1)−/− mice (back-crossed to the B6 strain for more than nine generations) were kindly provided by Dr. Gregg Duester (Sanford-Burnham Medical Research Institute, La Jolla, CA). The mice were bred in a specific pathogen-free facility (Bio Model System Park; KAIST, Daejeon, Korea). All animal experiments were approved by KAIST Institutional Animal Care and Use Committee. To induce liver fibrosis, 8- to 10-week-old mice were treated with 0.4 mL/kg carbon tetrachloride (CCl4) diluted in olive oil via intraperitoneal injection three times per week for 2 weeks. Twenty-four hours after the last injection of CCl4, 1 × 106 whole BMCs or control medium were transferred to mice via the tail vein. Twelve or 24 hours after infusion of BMCs, mice were sacrificed.

Human Study Population.

Human BMCs were harvested from patients with HBV-induced liver cirrhosis for autologous BMC infusion in Severance Hospital (Seoul, Korea). Some BMCs were used for in vitro experiments with the patients' consent. Serum data were obtained from patients treated with autologous BMC infusion between November 2006 and February 2008. The protocol for the clinical trial conformed to the ethical guidelines of the Declaration of Helsinki and was approved by the Institutional Review Board of Severance Hospital in the Yonsei University Health System.

Statistical Analysis.

Data are expressed as the mean ± SEM. To compare values, a Student t test or analysis of variance was performed. For blood samples of patients, a Wilcoxon signed-rank test with Bonferroni correction was used to compare the values of paired samples. P < 0.05 was considered statistically significant.

All other materials and methods are described in the Supporting Information.


Infused BMCs Ameliorate CCl4-Induced Liver Fibrosis in Mice.

To investigate early events following infusion of BMCs in fibrotic liver, mice with CCl4-induced liver fibrosis were sacrificed at 12 and 24 hours after infusion with BMCs from GFP+ mice via the tail vein. At 12 and 24 hours, serum levels of alanine aminotransferase, aspartate aminotransferase, triglyceride, albumin, cholesterol, and glucose were not changed compared with those of vehicle-infused mice (Fig. 1A and Supporting Fig.1A). However, collagen fibers and α-smooth muscle actin (α-SMA)–positive HSCs in liver tissues of BMC-infused mice were decreased compared with those of vehicle-infused mice at 12 and 24 hours (Fig. 1B and Supporting Fig. 1B), which were confirmed by western blotting (Fig. 1C) and quantitative RT-PCR (qRT-PCR) analyses (Fig. 1D) in isolated HSCs. In contrast, relative messenger RNA (mRNA) levels in HSCs for TGF-β1, IL-6, and monocyte chemoattractant protein-1 (MCP-1) were decreased only at 24 hours in BMC-infused mice but not in vehicle-infused mice, whereas there was no significant difference in IL-10 mRNA expression in HSCs (Supporting Fig. 1C). More surprisingly, most migrated GFP+ BMCs were in close contact with activated HSCs in the fibrotic septa within 24 hours (Fig. 1E and Supporting Fig. 1D). These results suggest that migrated BMCs might influence collagen production by activated HSCs via direct cell-cell interaction.

Figure 1.

Infused BMCs ameliorate CCl4-induced liver fibrosis in mice. Mice with 2-week CCl4-induced liver fibrosis were sacrificed at 12 or 24 hours after infusion of BMCs of GFP+ mice via the tail vein. (A) Serum chemistry was conducted for alanine aminotransferase (ALT) and albumin. (B) Representative sirius red or α-SMA staining of liver tissue at 24 hours after BMC infusion (original magnification, ×100). (C) Isolated HSCs of liver tissues were subjected to western blotting. (D) α-SMA and COL1A1 mRNA levels in isolated HSCs were determined via qRT-PCR. (E) After staining with α-SMA antibody, frozen liver sections were observed under fluorescence microscopy (original magnification: upper panels, ×100; lower panels, ×400). Data are expressed as the mean ± SEM (at least 8-12 mice per group). *P < 0.05, **P < 0.01 versus corresponding controls.

Infused BMCs Increase IL-10 Expression and CD4+CD25+Foxp3+ Tregs While Decreasing the Expression of IL-6 and MCP-1, and Reducing CD11b+F4/80+ Cells in Fibrotic Liver.

We next examined the changes of inflammatory mediators and cells in the liver after infusion of BMCs. Isolated liver mononuclear cells (MNCs) of BMC-infused mice had a higher expression of IL-10 and Foxp3 but a reduced expression of proinflammatory MCP-1 and IL-6 compared with vehicle-infused mice (Fig. 2A). Because Foxp3 is a master regulator of Tregs that induces production of IL-10, we analyzed intrahepatic frequencies of Tregs by fluorescence-activated cell sorting (FACS) analyses (Supporting Fig. 2A). By gating for liver lymphocytes, mice with infused BMCs displayed a significant increase in CD4+CD25+Foxp3+ Tregs compared with that of vehicle-infused mice, and the increased Tregs did not express GFP, suggesting that they were derived from recipient mice (Fig. 2B and Supporting Fig. 2B).

Figure 2.

Infused BMCs increase IL-10 expression and CD4+CD25+Foxp3+ Tregs while decreasing the expression of IL-6 and MCP-1, and reducing F4/80+CD11b+ cells in fibrotic liver. Mice with 2 weeks of CCl4-induced liver fibrosis were sacrificed at 12 or 24 hours after infusion of BMCs of GFP+ mice via tail vein, and liver MNCs and tissues were isolated or collected. (A) Isolated liver MNCs were subjected to qRT-PCR. (B, C) CD4+CD25+Foxp3+ Tregs and F4/80+CD11b+ macrophages in liver MNCs were analyzed at 24 hours after infusion of BMCs. (D, E) TGF-β1 mRNA levels of whole liver tissues and isolated liver MNCs were determined via qRT-PCR. Data are expressed as the mean ± SEM (at least 8 mice per group). *P < 0.05, **P < 0.01 versus corresponding controls.

Because the anti-inflammatory effects of Tregs are attributable to IL-10 and TGF-β1, we assessed the intrahepatic infiltration of CD11b+F4/80+ macrophages, NK1.1+CD3 NK cells, and Gr1+CD11b+ granulocytes. Whereas the numbers of NK cells and granulocytes were not affected by infusion of BMCs (Supporting Fig. 2C), CD11b+F4/80+ macrophages were significantly decreased at 24 hours after BMC infusion compared with those of vehicle-infused mice (Fig. 2C and Supporting Fig. 2D). In addition, many of GFP+ BMCs were observed in the regions where there were decreased numbers of CD11b+F4/80+ cells (Supporting Fig. 2D). Some of the infused BMCs were double-positive for GFP (green) and F4/80 (red) in the inflammatory regions (Supporting Fig. 2E). Because TGF-β1 is not only an important driver of liver fibrosis but also a major cytokine of Tregs, macrophages, and HSCs, we assayed TGF-β1 expression in whole liver tissues, isolated HSCs and liver MNCs. TGF-β1 expression in whole liver tissues and isolated HSCs was ameliorated in BMC-infused mice compared with those of vehicle-infused mice at 24 hours (Fig. 2D and Supporting Fig. 1C). In contrast, TGF-β1 expression in liver MNCs was significantly increased at 12 hours, whereas there was no difference at 24 hours in BMC-infused compared with vehicle-infused mice (Fig. 2E).

Infused BMC Subtypes CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+ Produce IL-10 in Fibrotic Livers of Recipient Mice.

Similar to a previous report,16 approximately 0.3% of liver MNCs in recipient mice were composed of GFP+ BMCs at 12 and 24 hours after BMC infusion (Fig. 3A). Moreover, less than 0.1% and 0.6-1.0% GFP+ cells were identified in gates of lymphocytes and monocyte/granulocytes, respectively (Fig. 3A). Furthermore, in analyzing infused BMCs in fibrotic liver, almost all GFP+ cells had originated from bone marrow–derived hematopoietic cells (CD45-positive), and most of them (75%-80%) expressed CD11b and Gr1 (Supporting Fig. 3A), which are specific markers for the myeloid-cell lineage differentiation.4, 17 Thus, we analyzed GFP+ BMCs using antibodies to Gr1 and F4/80 to distinguish between granulocyte and monocyte lineages. Around 75% of GFP+ BMCs were positive for Gr1, whereas 20% of them expressed F4/80 (Fig. 3B). Next, we investigated subsets of infused BMCs using antibodies to CD11b, Gr1, and F4/80 after gating with CD45. Most of the GFP+ BMCs (≈80%) were double-positive for Gr1 and CD11b, and after gating with CD45 and CD11b, CD11b+Gr1highF4/80, CD11b+Gr1lowF4/80, and CD11b+Gr1+F4/80+ cells comprised about 25%, 16%, and 15% of infused GFP+ BMCs, respectively (Supporting Fig. 3). More surprisingly, IL-10–positive infused BMCs were identified as CD11b+Gr1+ cells, which could be further subdivided into CD11b+Gr1highF4/80, and CD11b+Gr1+F4/80+ cells (Fig. 3C,D). Thus, IL-10+CD11b+Gr1+F4/80+ and IL-10+CD11b+Gr1highF4/80 BMCs appear to be undifferentiated cells that might belong to the monocytic and granulocytic lineages, respectively, based on their morphology, cytoplasmic granules, and CD markers (Fig. 3C-E).

Figure 3.

Infused BMC subtypes CD11b+Gr1+F4/80+ and CD11b+Gr1highF4/80 play important roles in IL-10 production in fibrotic livers of recipient mice. Mice with 2-week CCl4-induced liver fibrosis were sacrificed at 12 or 24 hours after infusion of BMCs of GFP+ or CD45.1+ mice via the tail vein. Liver MNCs were isolated and subjected to FACS analyses. (A) The percentages of infused BMCs (GFP+) to total liver MNCs were analyzed. (B) The cellular markers of GFP+ BMCs were analyzed using antibodies to CD45, CD11b, Gr1, and F4/80. (C, D) Intracellular cytokine staining of IL-10 was performed and analyzed for IL-10+CD45.1+ BMCs of isolate liver MNCs. (E) Among IL-10+CD45.1+ BMCs of isolate liver MNCs, CD11b+Gr1+F4/80+ (top panel) and CD11b+Gr1highF4/80 (bottom panel) BMCs were sorted and stained with Giemsa (original magnification, ×2000). Data are expressed as the mean ± SEM (at least 5-10 mice per group). *P < 0.05, **P < 0.01 versus corresponding controls.

Coculturing with HSCs Enhances IL-10 Expression by BMCs, Which Suppresses Expression of α-SMA and Type 1 Collagen alpha 1 in HSCs.

Because infused BMCs in the fibrotic area were adjacent to activated HSCs and displayed increased IL-10 expression (Figs. 1E and 3C), we hypothesized that enhanced IL-10 expression in infused BMCs might be due to their interactions with HSCs. To test this hypothesis, we cocultured BMCs with activated HSCs up to 24 hours (Fig. 4A and Supporting Fig. 4A). IL-10 expression in adherent and floating BMCs significantly increased after coculturing, but floating BMCs expressed higher IL-10 than adherent BMCs at 6 hours (Fig. 4B and Supporting Fig. 4B). In contrast, expression of α-SMA and type 1 collagen alpha 1 (COL1A1) genes in HSCs was significantly reduced by coculturing with BMCs (Fig. 4C).

Figure 4.

Coculturing with HSCs enhances IL-10 expression by BMCs, which suppresses activation of D4 HSCs. (A) Isolated HSCs from healthy mice were cultured on a plastic dish for 4 days (D4 HSCs). Freshly isolated BMCs were then cocultured with activated D4 HSCs for 6-24 hours (HSCs/BMCs = 1:10). (B) IL-10 mRNA levels in adherent and floating BMCs of mice were determined via qRT-PCR. (C) After coculturing, α-SMA and COL1A1 mRNA levels of D4 HSCs of mice were determined via qRT-PCR. (D) Human HSCs (LX-2 and hTERT cell lines) were cocultured with human BMCs of hepatitis B virus patients for 6-24 hours after 6-hour starvation of human HSCs. IL-10 mRNA levels in floating human BMCs were determined via qRT-PCR. (E) IL-10 levels of sera in hepatitis B virus patients (n = 15) with cirrhosis were measured at various time points and analyzed after autologous BMC infusion. (F) After autologous BMC infusion, IL-10 levels in blood were analyzed in the effective patients (n = 10) of all patients (n = 15) showing improvement of Child-Pugh score and albumin levels. Data are expressed as the mean ± SEM for 3-5 independent experiments. *P < 0.05, **P < 0.01 versus corresponding controls.

Next, we examined whether IL-10 secretion from human BMCs could be enhanced by coculturing with human HSCs (Supporting Fig. 4C). Once human BMCs stuck to HSCs, it was difficult to separate the two cell types; therefore, we collected only floating human BMCs after coculturing and analyzed expression of IL-10. In qRT-PCR analyses, IL-10 expression was increased in human BMCs cocultured with LX-2 and hTERT HSC cell lines at 6 and 12 hours (Fig. 4D). These data were concordant with those of mice. Therefore, we assessed the IL-10 levels in the sera of patients (n = 15) with liver cirrhosis after autologous BMC infusion therapy. Patient information is provided in Supporting Table 1. After autologous BMC infusion, a trend toward increased IL-10 was detected in the sera of patients, which was not statistically significant by Bonferroni correction (Fig. 4E). We further analyzed IL-10 levels in patients. First, patients were separated into two groups as follows: After autologous BMC infusion, patients with improved Child-Pugh scores and albumin levels (n = 10) were designated as the effective group and patients with no improvements (n = 5) were designated as the noneffective group. Surprisingly, patients in the effective group after autologous BMC infusion expressed significantly more IL-10 at day 1 (P = 0.03027 by Bonferroni correction), which was sustained for 14 days (not significant), whereas patients in the noneffective group had no difference in IL-10 levels compared with those of day 0 (Fig. 4F and Supporting Fig. 4D). These data reinforce IL-10 as a potential factor in the early response to BMC infusion therapy for treatment of hepatic fibrosis in mice as well as humans.

CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+ BMC Enhance IL-10 Expression After Coculturing with D4 HSCs.

To further investigate IL-10 expression by BMCs in vitro, we analyzed the subsets of BMCs after coculturing with HSCs. Since the major sources of IL-10 among infused BMCs were identified as CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+ cells in vivo (Fig. 3C), we investigated whether adherent and floating BMCs contained both types of cells. In FACS analyses after coculturing, adherent BMCs contained a higher fraction of CD11b+Gr1+F4/80+ cells (18%) than those of floating cells (6%), while the frequency of CD11b+Gr1highF4/80 cells (87%) in floating BMCs exceeded that of adherent cells (50%) at 6 hours (Fig. 5A and Supporting Fig. 5A). After 6 hours of coculture, IL-10–positive cells in adherent and floating BMCs were higher than those of control BMCs, respectively (Fig. 5B and Supporting Fig. 5B). Therefore, we further analyzed IL-10–positive cells of BMCs using antibodies to CD11b, Gr1, and F4/80. After coculturing with HSCs, the frequencies of CD11b+IL-10+ cells in adherent (8%) and floating (5%) BMCs were much higher than those (4.7% and 1.8%) of control BMCs; CD11b+Gr1+F4/80+ cells and CD11b+Gr1highF4/80 cells were identified as major IL-10–producing cells in adherent and floating BMCs, respectively (Fig. 5C,D and Supporting Fig. 5C). However, CD11bIL-10+ cells in control and cocultured BMCs showed similar frequencies, which were mostly recognized as CD11bGr1+F4/80+ cells (Supporting Fig. 5D).

Figure 5.

CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+ BMCs enhance IL-10 expression after coculturing with D4 HSCs. BMCs isolated from normal mice were cocultured with D4 HSCs. (A) Adherent and floating BMCs were analyzed using antibodies to Gr1 and F4/80 after coculturing with D4 HSCs for 6 hours. (B) IL-10–positive BMCs increased in both adherent and floating BMCs after coculturing. (C, D) IL-10–positive BMCs were analyzed using antibodies to IL-10, CD11b, Gr1, and F4/80. (E) Cocultured BMCs (CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+) were sorted using FACS Aria and then stained with Giemsa or intracellular cytokine staining of IL-10 after further culturing with monensin for an additional 18 hours (original magnification, ×1200). IL-10–positive cells are indicated with a black arrowhead or black arrow; IL-10–negative cells are indicated with a white arrowhead or white arrow. Data are expressed as the mean ± SEM for 5 independent experiments. *P < 0.05, **P < 0.01 versus corresponding controls.

To characterize the morphologies of IL-10–producing BMCs, CD11b+Gr1+F4/80+ and CD11b+Gr1highF4/80 cells were sorted and then stained with Giemsa followed by immunocytochemistry for IL-10. Using Giemsa staining, monocytic cells with vesicles and granules were the major types among the CD11b+Gr1+F4/80+ adherent BMCs, in which monocytic cells with nonindented nuclei were positive for IL-10 (Fig. 5E, upper panels). In contrast, granulocytic cells and their precursor cells were the main cell types among CD11b+Gr1highF4/80 floating BMCs, in which precursor type cells were positive for IL-10 (Fig. 5E, lower panels). In addition, in further analyses of BMCs with additional antibodies to Ly6G and Ly6C, the CD11b+Gr1+F4/80+ and CD11b+Gr1highF4/80 cells were identified as CD11b+Ly6GLy6Chigh and CD11b+Ly6G+Ly6Clow cells, respectively (Supporting Fig. 5E). Based on these findings, adherent and floating BMCs expressing IL-10 might be monocytic and granulocytic MDSC-like cells, respectively. Other Gr1lowF4/80 BMCs were identified as precursor cells for granulocytes and monocytes (Supporting Fig. 5F).

Infused BMC-Derived IL-10 Is a Key Molecule for Ameliorating Liver Fibrosis and Expansion of Liver Tregs.

To confirm the antifibrotic role of infused BMC-derived IL-10 in liver fibrosis, we infused IL-10–deficient BMCs in mice with CCl4-induced liver fibrosis. As expected, sirius red staining and immunohistochemistry for α-SMA demonstrated that liver fibrosis was significantly ameliorated in wild-type (WT) BMC-infused mice but not in IL-10–deficient BMC-infused mice compared with controls (Fig. 6A). Image analyses and western blotting further confirmed these findings (Fig. 6B,C). qRT-PCR showed significantly increased IL-10 mRNA expression in liver MNCs of WT BMC-infused mice, but not in IL-10–deficient BMC-infused mice compared with controls (Fig. 6D). Moreover, frequencies of Tregs in livers of IL-10–deficient BMC-infused mice were unchanged compared with controls (Fig. 6E,F). These data indicate that infused BMC-derived IL-10 is a key molecule that accounts for the antifibrotic activity observed in this model.

Figure 6.

Depletion of the IL-10 gene in BMC abrogates the antifibrotic effects of BMC via down-regulation of IL-10 expression and Treg induction in liver MNCs of recipient mice. Mice with 2-week CCl4-induced liver fibrosis were infused with vehicle and isolated whole BMCs of control WT or IL-10−/− mice via the tail vein. Mice were sacrificed at 12 and 24 hours after BMC infusion. (A, B) To estimate liver fibrosis and HSC activation, sections of liver tissues were stained with sirius red and α-SMA antibody. The stained areas of sirius red and α-SMA antibody were calculated with image analyzer program. (C) Whole liver tissues were subjected to western blotting. (D) IL-10 mRNA levels in liver MNCs were determined via qRT-PCR analysis. (E, F) Intrahepatic CD4+CD25+FoxP3+ Tregs increased in the mice treated with WT BMCs but not in IL-10−/− BMCs at 24 hours. Data are expressed as the mean ± SEM (at least 8-10 mice per group). *P < 0.05 versus corresponding controls.

IL-10 Expression in BMCs Is Promoted or Suppressed by Retinoic Acid and IL-6 of HSCs, Respectively.

Finally, we sought to identify mediators of HSCs that affected expression of IL-10 in BMCs. Because HSCs can produce IL-6, IL-10, and RALDH1-mediated retinoic acid, these factors have been considered as candidate components driving the inflammatory reaction, and expansion and differentiation of Tregs and MDSCs.11, 18-21 Accordingly, we cocultured BMCs with IL-6, IL-10, and RALDH1 gene-depleted HSCs, respectively. In the absence of IL-6 in HSCs, IL-10 expression was significantly increased in both adherent and floating BMCs compared with those of WT BMCs cocultured with WT HSCs (P < 0.05), whereas RALDH1-deficient HSCs did not increase IL-10 expression by BMCs compared with those of WT BMCs cocultured with WT HSCs (Fig. 7A, B). In addition, IL-10–deficient WT HSCs increased IL-10 expression similarly in both adherent and floating BMCs compared with those of WT BMCs cocultured with WT HSCs (Fig. 7A,B). To reinforce the effect of retinoic acid on IL-10 production by infused BMCs in vivo, we administrated CCl4 to RALDH1-deficient mice for 2 weeks, and these animals were then infused with WT BMCs. Twenty-four hours after infusion of BMCs, fibrosis was not ameliorated (Fig. 7C and Supporting Fig. 6A). Based on FACS analyses, there were no significant changes in the frequencies of inflammatory cells, such as CD11b+F4/80+ macrophages and CD11b+Gr1+ granulocytes, and Tregs as well in liver (Fig. 7D and Supporting Fig. 6B,C).

Figure 7.

IL-10 expression of BMCs is promoted or suppressed by retinoic acid and IL-6 of HSCs, respectively. Isolated HSCs from WT, IL-6−/−, IL-10−/−, and RALDH1−/− mice were cultured on plastic plates for 4 days (D4 HSCs). Freshly isolated WT BMCs from control mice were then cocultured with activated D4 HSCs for 6-24 hours (HSCs/BMCs = 1:10) or cultured alone as control. (A, B) After coculturing, IL-10 expression in the adherent and floating BMCs was analyzed via qRT-PCR. RALDH1−/− mice with 2-week CCl4-induced liver fibrosis were infused with isolated whole BMCs of WT mice via the tail vein and then were sacrificed at 24 hours after infusion of BMCs. (C) Representative sirius red staining of liver tissue (original magnification ×100). (D) There were no significant changes in the population of Gr1+CD11b+ and F4/80+CD11b+ cells in liver, and intrahepatic CD4+CD25+FoxP3+ Tregs did not increase in RALDH1−/− mice after infusion of WT BMCs at 24 hours. (E) An integrated depiction of antifibrotic roles of infused BMCs after interaction with HSCs. First, the interaction between infused BMCs and HSCs is crucial for IL-10 production in infused BMCs (especially CD11b+Gr1+F4/80+ and CD11b+Gr1highF4/80 MDSC-like cells). Second, retinoic acid and IL-6 from activated HSCs stimulate and inhibit, respectively, the production of IL-10 by infused BMCs. Third, IL-10 from infused BMCs of donor mice is essential for the induction of Tregs in recipient mice, which then suppresses activated HSCs via direct (IL-10 production) and indirect (Treg induction) effects of infused BMCs. Data are expressed as the mean ± SEM for three independent experiments or using 8 mice per group. *P < 0.05, **P < 0.01 versus corresponding controls.


The beneficial effects of BMC therapy have been investigated recently in mice and humans, yet underlying mechanisms have been overlooked, especially the early effects of BMCs. In the present study, we identify early phase antifibrotic effects of infused BMC in vivo and in vitro, which reflect the interaction between HSCs and BMCs within 24 hours. The mechanisms of liver fibrosis amelioration by infused BMCs are summarized in Fig. 7E.

Contrary to the reported long-term effects of BMCs in fibrotic livers of mice and humans,1-3 we have shown that at early time points, infused BMCs ameliorate liver fibrosis without any change in liver injury, hepatocyte regeneration, or albumin production (Fig. 1 and Supporting Fig. 1A), suggesting that there are no effects of bone marrow–derived stem cells within 24 hours after infusion. In addition, the involvement of BMC-derived myofibroblasts in collagen production appears to be negligible during liver fibrogenesis.22 Instead, we found that infused GFP+ BMCs, mostly expressing CD11b and Gr1, migrated into fibrotic septa adjacent to activated HSCs (Fig. 1E and Supporting Fig. 3A), which might result from the expression of CCR2 and MCP-1 in BMCs and HSCs, respectively,5, 23, 24 whereas isolated HSCs of liver after infusion of BMCs have decreased expression of α-SMA, COL1A1, TGF-β, IL-6, and MCP-1 genes compared with those of controls (Fig. 1D and Supporting Fig. 1C). These findings suggest that infused BMCs interact with HSCs and suppress liver fibrosis by inhibiting their activation.

In the present study, the expression of IL-10 was significantly increased in liver MNCs within 24 hours following infusion of BMCs, and its antifibrotic effect was abrogated when IL-10–deficient BMCs were infused instead. Moreover, IL-10 expression of BMCs was enhanced by coculturing with activated HSCs, whereas activation of HSCs was inversely related to IL-10 expression (Fig. 4B,C). These coculture findings are especially informative for the following reasons. First, both adherent and floating BMCs contained CD11b+Gr1+F4/80+ and CD11b+Gr1highF4/80 cells (Fig. 5A). Second, although both CD11b+Gr1+F4/80+ and CD11b+Gr1highF4/80 cells were enriched in adherent BMCs and floating BMCs, respectively, these distributions changed over time. For instance, the population of CD11b+Gr1highF4/80 cells in floating BMCs decreased slowly after coculturing, whereas their representation within adherent BMCs increased, and then a similar fraction was detected in adherent and floating BMCs at 24 hours (Supporting Fig. 5A, left panel). Moreover, the population of CD11b+Gr1+F4/80+ cells in adherent BMCs slowly decreased and then approximated those of floating BMCs at 24 hours after coculturing with HSCs (Supporting Fig. 5A, right panel). Therefore, it is unclear whether there are differences between the adherent and nonadherent BMCs based on the coculture experiments. Further studies will be needed to resolve this question. In parallel to the murine data, enhanced expression and production of IL-10 were confirmed within 24 hours of coculturing human BMCs with human HSC lines (Fig. 4D) and in sera of human patients, respectively, consistent with the beneficial effects following autologous BMC infusion (Fig. 4F). These findings indicate that BMC production of IL-10 is not only a critical event at an early phase after infusion of BMCs, but is also a crucial negative regulator of liver fibrosis, as reported.5, 6 Indeed, the source of IL-10 is primarily from infused BMCs, especially CD11b+Gr1highF4/80 and CD11b+Gr1+F4/80+ cells (Fig. 3C), and these cells were also identified as CD11b+Ly6G+Ly6Clow and CD11b+Ly6GLy6Chigh cells, respectively (Supporting Fig. 5E).

Based on prior reports, MDSCs are a heterogeneous population of immature cells capable of inhibiting immune responses, and were originally characterized based on their coexpression of the myeloid-cell lineage differentiation marker Gr1 and CD11b.12, 17, 20 To date, MDSCs are distinguished between two subsets: granulocytic MDSCs have a CD11b+Ly6G+Ly6Clow phenotype, whereas monocytic MDSCs have a CD11b+Ly6GLy6Chigh phenotype.17 Thus, IL-10+ BMCs detected in recipient mice share the same markers with MDSCs, as specific cells with a nonlobulated nucleus that produce IL-10 (Figs. 3E and 5E). Moreover, recent studies demonstrate that HSCs can promote generation of MDSCs in vivo and in vitro, thereby protecting islet allografts against immune cell attack.12 MDSCs can also increase IL-10 production after cell-cell contact with macrophages of tumor-bearing mice.25 These studies support our results that infiltrated BMCs in fibrotic liver express the same makers as MDSCs, and they further increase IL-10 expression after interacting with activated HSCs. In addition, we found an increased population of CD4+CD25+Foxp3+ Tregs originating from recipient mice after infusion of BMCs that are also anti-inflammatory based on their production of IL-10 and TGF-β (Fig. 2B).15, 18 According to recent studies, MDSCs of patients and mice with tumors contribute to the induction of Tregs.13, 14, 17, 26 Treg induction also requires IL-10 and TGF-β of MDSCs,14 which preferentially induces proliferation of natural Tregs26 leading to reduced activation of macrophages and T cells. In our study, enhanced IL-10 production of infused BMCs decreased the population of macrophages (Fig. 2C and Supporting Fig. 2D) and expanded Tregs in liver MNCs of recipient mice, which was reversed in recipient mice after infusion of IL-10–deficient BMC (Fig. 6D-F).

According to previous studies, TGF-β, IL-6, and retinoic acid are not only important factors in T cell differentiation8 but also in the activation and further differentiation of MDSCs into macrophages, dendritic cells, and granulocytes.14, 19-21 Intriguingly, HSCs can produce a variety of mediators, including TGF-β, IL-6, and retinoic acid, depending on their state of activation.5 Thus, to clarify which mediators of HSCs play an important role in BMC production of IL-10, we cocultured BMCs with HSCs deficient in the production of IL-10, IL-6, and RALDH1 or WT HSCs (Fig. 7A,B). Surprisingly, IL-6–deficient HSCs induced more IL-10 expression by BMCs, whereas RALDH1-deficient HSCs had decreased IL-10 compared with that of BMCs cocultured with WT HSCs. Moreover, RALDH1-deficient mice displayed decreased production of retinoic acid27 and did not show any antifibrotic effects of infused WT BMCs (Fig. 7C,D and Supporting Fig. 6A). However, IL-10–deficient HSCs did not affect production of IL-10 by WT BMCs. Thus, retinoic acid metabolized from retinol by RALDH1 and IL-6 in HSCs might play important roles in IL-10 production by BMCs. In support of this prospect, treatment with retinoic acid increased IL-10 production in several cell lines, possibly because the IL-10 locus harbors at least one retinoic acid–response element.28 In addition, the antagonist effect of retinoic acid on IL-6 has been demonstrated during T cell differentiation.8 However the underlying mechanism is not clear. Therefore, further studies are necessary to understand the roles of IL-6 and retinoic acid in the production of IL-10 in BMCs.

Recently, intriguing studies suggest that different subsets of macrophages and dendritic cells have varying roles in liver injury, fibrosis, and tumor development. For instance, delivery of bone marrow–derived macrophages differentiated by colony-stimulating factor-1 are reportedly beneficial by reducing fibrosis, mainly by recruiting endogenous macrophages and neutrophils producing matrix metalloproteinase (MMP)-9 and MMP-13; however, this protective effect was not detected in treatment of macrophage precursors.29 In our study, MMP-9 and MMP-13 were increased in isolated liver MNCs from both WT and IL-10–deficient BMC-treated mice compared with those of controls (Supporting Fig. 7B,C). Thus, the expression of MMPs was not a critical determinant of the findings in our study. In addition, dendritic cells can reduce liver ischemia/reperfusion injury and fibrosis via IL-10 secretion and MMP-9 expression, respectively.30, 31 In contrast, CD11b+F4/80+Gr1+ macrophages promote liver fibrosis and tumor development in a TGF-β–dependent manner.32, 33 These reports also demonstrate that various types of bone marrow–derived cells acquire different functions during liver injury. Thus, further studies to characterize functional subsets of bone marrow–derived cells should be pursued.

In conclusion, we provide evidence in mice and humans that IL-10 production by infused BMCs is a key negative regulator of liver fibrosis at early time points. Crucially, the interplay between HSCs and BMCs is necessary for the induction of IL-10 in infused BMCs (CD11b+Gr1+F4/80+ and CD11b+Gr1highF4/80 MDSC-like cells), which in turn expand the Treg population in recipient mice. Our findings may contribute to the refinement of autologous BMC therapeutic approaches for patients with liver fibrosis and cirrhosis.