Bone marrow transplantation demonstrates medullar origin of CD34+ fibrocytes and ameliorates hepatic fibrosis in Abcb4−/− mice


  • Potential conflict of interest: Nothing to report.


Bone marrow (BM)-derived stem cells and CD34+ fibrocytes are associated with fibrogenesis in several organs. In an Abcb4−/− mouse model for sclerosing cholangitis alpha-smooth muscle actin-positive (α-SMA+) myofibroblasts are thought to play a pivotal role in hepatic fibrogenesis. The aim of this study was 2-fold: (1) to demonstrate that the origin of an important fibrogenetic cell population is the BM; and (2) to investigate whether transplantation of BM (BM-Tx) affects liver function, staging, and grading. Surrogate markers for fibrogenesis and regulation of hepatic stellate cells (HSC) as well as progenitor-cell-derived fibrocytes in liver tissue were analyzed by quantitative real-time polymerase chain reaction (PCR) and immunohistology. After lethal irradiation of recipient mice, BM-Tx was carried out by way of tail vein injection of BM cells from marker protein donors (green fluorescent protein, GFP+) or Abcb4−/− mice as control (syngeneic Tx). Parameters of liver function were assessed serologically and histologically. Activated HSC of α-SMA+/CRP2+ phenotype were expressed in ≈50% of proliferating bile ducts, whereas fibrotic liver parenchyma showed no expression thereof. Epithelial mesenchymal transfer (EMT) was visualized in the areas of proliferating bile ducts. The hematopoietic origin of CD34+ fibrocytes was demonstrated immunohistologically in livers of BM chimeric mice. These CD34+ cells infiltrated hepatic lobules from portal fields and developed a desmin+ phenotype expressing collagen type I in fibrotic parenchyma as well as in vitro after isolation by magnetic cell separation. Transplantation of GFP+/Abcb4+ BM improved liver function and staging compared with sham transplantation, but no significant differences were noticed among allogeneic and syngeneic Tx. Conclusion: The present study is the first to identify that both BM-derived fibrocytes and HSC are involved in biliary fibrogenesis in Abcb4−/− mice. Our data suggest that changes in immunity subsequent to BM-Tx may alter hepatic fibrosis. (HEPATOLOGY 2009.)

Sclerosing cholangitis comprises a spectrum of chronic biliary diseases that has either an unknown etiology (i.e., primary) or is caused by identifiable insults to the biliary tree (i.e., secondary).1 Based on a prevalence of 13.9 per 100,000, it is calculated that ≈29,000 individuals across the USA suffer from primary sclerosing cholangitis (PSC).2 PSC may emerge unnoticed with a rather unspecific clinical presentation and large variability in clinical courses. The pathology of PSC is characterized by progressive obliteration of the bile ducts with subsequent biliary cirrhosis, eventually leading to portal hypertension and liver failure.

Abcb4 (an ortholog of human MDR3/ABCB4) knockout mice represent a highly reproducible, well-characterized nonsurgical mouse model for cholangiopathy, clearly showing the macroscopic (bile duct strictures and dilatations of the large bile ducts) and microscopic features (onion skin-type like pericholangitis and periductal fibrosis) of sclerosing cholangitis in humans.3, 4

Mutations of the human ABCB4 gene result in phenotypes of chronic liver disease like progressive familial intrahepatic cholestasis (PFIC type 3) or biliary liver cirrhosis.5 Physiologically, the ABCB4 gene encodes the P-glycoprotein ABCB4, a member of the adenosine triphosphate (ATP) binding cassette (ABC) transporters, which is responsible for phospholipid transport across the canalicular membrane. The lack of this transporter in Abcb4−/− mice causes the complete absence of micelle-forming phospholipids from bile, which results in liver injury from chronic cholangitis.3, 6

Although PSC has a close association with ulcerative colitis, the etiology of the disease remains unclear. Transdifferentiation of hepatic stellate cells (HSC) into α-smooth muscle actin (α-SMA) expressing myofibroblasts and subsequent deposition of extracellular matrix (above all, collagen type I) is thought to be a cellular key mechanism for the development of fibrosis in PSC.7

Fibrocytes constitute a circulating bone marrow (BM)-derived CD34+ cell population with fibroblast-like properties initially associated with tissue repair in subcutaneous wounds.8 They comprise a fraction below 1% of the circulating pool of leukocytes with markers of mesenchymal cells.8, 9 In particular, they are involved in organ fibrosis, e.g., pulmonary fibrosis, kidney fibrosis, and also dermal fibrosis.9–13 Hepatic myofibroblasts originating from BM were first identified in 2004 by Forbes et al.14 by way of Y chromosome tracking in a female patient who developed cirrhosis after receiving a BM transplant from a male. Subsequently, two studies functionally demonstrated BM origin of fibrotic cell populations in mouse models of carbon tetrachloride (CCl4) intoxication and bile duct ligation.15, 16

Just recently, it has been presumed that bile duct epithelial cell-derived growth factors may activate and recruit BM-derived progenitor cells and transform to fibrocytes, finally promoting fibrogenesis in sclerosing cholangitis.4

Against this background, we hypothesized that BM-derived circulating CD34+ fibrocytes represent key mediators of liver fibrogenesis in Abcb4−/− mice. Syngeneic and allogeneic BM-transplantation (BM-Tx) was conducted (1) to demonstrate that the origin of the fibrogenetic cell population is indeed the BM; and (2) to investigate whether transplantation of BM from Abcb4+/+ into Abcb4−/− mice may ameliorate liver function characterized by liver staging and grading.


α-SMA; alpha-smooth muscle actin; ABCB4, ATP binding cassette (transporter, family member) B4; ALT, alanine aminotransferase; AP, alkaline phosphatase; AST, aspartate aminotransferase; ATP, adenosine triphosphate; BM; bone marrow; BM-Tx bone marrow transplantation; CRP-2, cysteine and glycine rich protein-2; EMT, epithelial mesenchymal transfer; FACS, fluorescence-activated cell sorting; GFP, green fluorescent protein; HSC, hepatic stellate cell; HYP, hydroxyproline; MACS, magnetic-activated cell sorting; MMP-9, matrix metalloproteinase-9; PSC, primary sclerosing cholangitis; SDF-1, stromal cell-derived factor-1.

Materials and Methods

Animal Model.

The present study was performed with permission of the State of Hesse, Regierungspräsidium Giessen, according to Section 8 of the German Law for the Protection of Animals and conforms to the NIH Guide for the Care and Use of Laboratory Animals. BALB/c-Abcb4−/− mice were bred and housed as described.17 Progressive liver injury develops spontaneously and resembles human sclerosing cholangitis.6 Confirmation of the Abcb4−/− genotype of knockout mice by polymerase chain reaction (PCR) was demonstrated.18 At age 4, 6, 8, 10, 12, 14, 16, and 28 weeks mice were killed by CO2-inhalation (n = 5 per age and sex). Liver samples were collected and preserved for analyses, as indicated. For routine biochemistry, serum samples were stored at −85°C until analysis of aspartate and alanine aminotransferases (AST, ALT) and alkaline phosphatase (AP) by routine clinical chemistry on a Reflotron Plus Analyzer (Roche, Mannheim, Germany).

BM Transplantation.

BALB/c-GFP transgenic mice were raised from C57BL/6-TgN(ACTbEGFP)1Osb (Jackson Laboratories, Bar Harbor, ME) and crossed back on BALB/c for 10 generations. The mice were kindly provided by Dr. M. Heil (Max-Planck-Institute, Bad Nauheim, Germany). Six-week-old BALB/c-Abcb4−/− (n = 8) were lethally irradiated (11 Gy, 60Co) and 5*106 BM donor cells from BALB/c-GFP transgenic littermates or from BALB/c-Abcb4−/− for syngeneic control were transplanted by way of tail vein injection. Owing to the GFP-marker, all of the tissues of GFP transgenic mice, except erythrocytes and hair follicle cells, appear green under blue excitation light.19 A second control group (n = 8) received an equal volume of phosphate-buffered saline (PBS; i.e., sham Tx) without prior irradiation. Controls and BM-Tx mice were housed for another 10 weeks after BM-Tx. The amount of GFP+ BM engraftment was evaluated by flow cytometry of peripheral blood lymphocytes using FACSCalibur (Becton Dickinson, Heidelberg, Germany) as described.20

Histopathology and Hydroxyproline Assay.

Immediately after necropsy, liver samples for histopathological evaluation were fixed in 4% neutral buffered paraformaldehyde at 4°C for 16 hours and embedded in paraffin. Paraffin sections (6 μm) were stained with hematoxylin-eosin (H&E) or Masson Trichrome for the detection of collagen fibers.17 Histomorphometric analysis of inflammatory stage and hepatic fibrosis grade were independently assessed by a pathologist and a trained scientist (C.D. and M.R.). Staging and grading followed semiquantitative scores that have been described.21, 22 The entire content of collagen was determined by hydroxyproline (HYP) quantification.23

Immunohistological and Immunocytological Analysis.

Four-μm frozen tissue sections or isolated fibrocytes on glass coverslips were fixed in acetone/methanol for 2 minutes at −20°C, washed with PBS, and subsequently blocked for 30 minutes with 5% bovine serum albumin and 0.1% cold fish skin gelatin (Sigma-Aldrich, Munich, Germany) in PBS with 0.1% Triton (Roth, Karlsruhe, Germany) and 0.05% Tween 20 (Serva, Heidelberg, Germany).

Fluorescence immunostaining was performed using mouse anti-α-SMA antibodies (1:100, Progen, Heidelberg, Germany), rabbit anti-type I collagen (1:100, Rockland, Gilbertsville, PA), rabbit anti-CRP2 antibodies (1:50, kind gift from R. Weiskirchen, Aachen, Germany), rabbit anti-S100A4 antibodies (1:100, Dako, Glostrup, Denmark), mouse anti-cytokeratin-7 antibodies (CK-7, 1:100, Millipore, Schwalbach, Germany), rat anti-CD34 (1:50, BD, Heidelberg, Germany), mouse anti-desmin (1:50, Santa Cruz Biotech, Santa Cruz, CA), goat anti-prodomain of collagen type I (1:50, Santa Cruz Biotech), rabbit anti-prodomain of collagen type I (1:50, Santa Cruz Biotech), and goat anti-GFP antibodies (1:100, Rockland). Fluorescence immunostaining was analyzed using a fluorescence microscope (Leica DMRB, Wetzlar, Germany; camera: Nikon Coolpix 5400, Düsseldorf, Germany). Prior to use of primary mouse antibodies, cryosections were pretreated with the M.O.M. Kit (Vector Laboratories, Burlingame, CA). Costaining of nuclei was performed with DAPI (4′,6-diamidino-2-phenylindole dihydrochloride, Sigma-Aldrich). Specificity of all immunofluorescence stainings was proved using equally concentrated unspecific isotype immunoglobulin G (IgG) instead of primary antibodies for negative controls.

Isolation of CD34+ Fibrocytes from Mouse Liver.

Perfusion and digestion of liver matrix was performed using a modified protocol for HSC isolation from rat liver.24Abcb4−/− mice (16 weeks old) were killed by an overdose of isoflurane inhalation and the abdomen was opened. A cannula, 0.4 mm in diameter, was inserted into the portal vein and fixed using a steel clip. Vena cava was opened and subsequently the liver was rinsed in situ with HBBS (without Ca2+ and Mg2+) with a flow rate of 3 mL/min until the liver was free of blood. Accordingly, perfusion was continued with Pronase E solution for 8 minutes and Collagenase H solution for 20 minutes (flow rate of 3 mL/min) until the liver architecture was completely dissolved. All buffers were prepared and equilibrated for use as described.24 The liver was gently removed from the abdomen, placed into sterile plastic dishes, and rinsed with PBS. Afterwards the liver capsule was opened, tissue was torn into small pieces, and remaining tissue pieces were further dissected by rotating on a magnetic stirrer (200 rpm) for 10 minutes at 37°C. The cell suspension was filtered through sterile nylon meshes of 100 μm and 40 μm and subsequently centrifuged for 10 minutes at 450g and 4°C. The pellet was resuspended in a magnetic-activated cell sorting (MACS) buffer. All further steps were performed at 4°C.

MACS with Fc-blocking was performed according to the manufacturer's instructions (Miltenyi Biotech, Bergisch Gladbach, Germany). Endothelial cells were labeled with rat antimouse CD31 antibodies (Sanbio, Beutelsbach, Germany) and secondary goat antirat magnetic bead-linked antibodies (Miltenyi Biotech) and depleted from cell suspension in a first MACS-step. The remaining fibrocytes in the flowthrough were labeled with rat antimouse CD34 antibodies (BD) and secondary goat antirat magnetic bead-linked antibodies (Miltenyi) and enriched in a two-step MACS procedure using a so-called LS and an MS column, respectively.

Homogeneity of isolated CD34+/CD31 fibrocytes was analyzed by flow cytometry using rat antimouse CD34-PE-labeled antibodies or the appropriate PE-labeled isotype and 7AAD (eBioscience, San Diego, CA) to determine cell death. For immunocytochemistry, cells were cultured on glass coverslips in 24-well culture plates using Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS).

Western Blot.

Sample preparation, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and western blotting was performed as described.17 Immunodetection of α-SMA was realized using mouse anti-α-SMA antibodies diluted 1:2,000 in TNBS-N with 1% bovine serum albumin (BSA) and antitubulin antibodies (NeoMarkers, Thermo Fisher, Fremont, CA).

RNA Extraction, Complementary DNA Synthesis, and Quantitative Real-Time PCR (qRT-PCR).

Total liver RNA was isolated using RNeasy Kit (Qiagen, Hilden, Germany). One μg of RNA was utilized for the reverse transcriptase reaction (Omniscript, Qiagen) according to the manufacturer's instructions. Real-time PCR was performed using the Platinum SYBR Green qPCR Kit (Invitrogen, Karlsruhe, Germany) according to the manufacturer's protocol. Real-time PCR of each gene-specific primer pair was optimized prior to the experiment to confirm the absence of any nonspecific amplification product. Primers were purchased from Eurofins (Ebersberg, Germany) and primer sequences are presented in Table 1.

Table 1. SYBR Green Real-Time qPCR Primer Sequences
Target GenePrimerSequence
r18Sforward5′- gaa taa tgg aat agg acc gcg g -3′
 reverse5′- gga act acg acg gta tct gat c -3′
GFAPforward5′- cgt gtg gat ttg gag aga aag -3′
 reverse5′- gtg agg tct gca aac tta gac c -3′
CRP2forward5′- gct acg gaa aga agt atg gac c -3′
 reverse5′- ctc agt cag agt tgt aga ctc c -3′
CD34forward5′- cct acc aat gag tct gtt gag g -3′
 reverse5′- gaa gta gta ggc agt atg cca g -3′
Desminforward5′- ct agc tcg tatt gac ctg gag c -3′
 reverse5′- gt acc att cttc agc ctc aga g -3′
SDF-1forward5′- aac cag tca gcc tga gct ac-3′
 reverse5′- ggg tca atg cac act tgt ctg -3′
MMP-2forward5′- gga tac cca ttt gat ggc aag g -3′
 reverse5′- gaa gcc ata ctt gcc atc ctt c -3′
MMP-9forward5′- gtg ttc ccg ttc atc ttt gag g -3′
 reverse5′- gaa acc cca ctt ctt gtc agt g -3′
MMP-13forward5′- cag ttt ctt tat ggt cca ggc g -3′
 reverse5′- cat cca cat ggt tgg gaa gtt c -3′
TIMP-1forward5′- gct aaa ttc atg ggt tcc cca g -3′
 reverse5′- gag aaa gct ctt tgc tga gca g -3′

qRT-PCR was performed on the Mx3000P (Stratagene, La Jolla, CA) by using 3-stage program parameters as follows: (1) 10 minutes at 96°C, (2) 40 cycles of 10 seconds at 95°C, 30 seconds at 57°C, and 30 seconds at 73°C, (3) 10 minutes at 73°C. The specificity of the PCR was confirmed by examination of the dissociation reaction plot subsequent to qRT-PCR. PCR products were separated on a 1.5% TAE agarose gel and visualized by staining with ethidium bromide to confirm the appearance of a single band of the correct molecular size. qRT-PCR data were analyzed using the ΔΔCt model.25

Statistical Analysis.

Statistical analysis was performed with SPSS 17.0 software (Chicago, IL). Considering nonnormally distributed parameters nonparametric tests were applied (i.e., Mann-Whitney U-test and Spearman's rank test). Results are presented as mean ± standard deviation (SD). A two-sided P < 0.05 was considered significant. Interobserver variability of histopathologic grading and staging was assessed using kappa and weighted kappa for ordinal scales. The calculation of weighted kappa assumes that the categories are ordered and accounts for how far apart the two scorers are. Given a scale consisting of four subsequent levels, e.g., A, B, C, and D, homogeneity between two assessors is larger if one classifies a subject into group B and the other into group C, in contrast to an allocation into A and D.


Progressive Biliary Fibrosis in Mature Abcb4−/− Mice Is Not Associated with Myofibroblast Activation.

Figure 1A displays the logarithmic progression of collagen accumulation between the age of 4 and 28 weeks reaching up to 900 μg HYP per gram liver tissue. No gender differences were observed (Fig. 1A). Immunohistological analysis demonstrated extensive periportal and septal type I collagen deposition (Figs. 2A, 6A).

Figure 1.

Fibrogenesis in Abcb4−/− mice. (A) Entire collagen was assessed by measurement of HYP in murine liver. Open triangles indicate female animals, black quads indicate male animals, and rhombs depict mean of gender values. Displayed is the logarithmic progression of collagen accumulation between the age of 4 and 28 weeks in contrast to wildtype controls. No significant gender differences were observed. (B) Overall hepatic α-SMA expression was evaluated by western blots. Until the age of 16 weeks, no significant change in hepatic α-SMA expression was observed in Abcb4−/− mice. All data were normalized to tubulin content. Fold increase to wildtype animals is indicated by the bar chart (mean ± SD, n = 5).

Figure 2.

α-SMA+ myofibroblasts adjacent to proliferating bile ducts in Abcb4−/− mice: evidence for HSC-activation. (A,B) Coimmunofluorescence stainings of α-SMA+ (green) and type I collagen (red) showed α-SMA+ cells around blood vessels. Areas with periportal and septal fibrosis are negative for α-SMA staining (A, scale bar = 80 μm, original magnification ×100; enlarged panel B, scale bar = 10 μm, original magnification ×1,000). (C,D) Coimmunofluorescence stainings of α-SMA+ (green) and HSC-activation marker CRP-2 (red) illustrate the participation of HSC-derived myofibroblasts in periductular fibrogenesis (C, scale bar = 40 μm, original magnification ×200; enlarged panel D, scale bar = 10 μm, original magnification ×1,000). Representative micrographs are shown. (E) Hepatic gene regulation of GFAP and CRP-2-specific HSC genes. The unregulated transcription of GFAP, a marker of quiescent and activated HSC, demonstrated a lack of HSC proliferation, whereas the 1.4-fold induction of CRP-2 suggests an activation of HSC. All data were normalized to r18S. Fold increase to wildtype animals is depicted (mean ± SD, n = 5); all mice were 16 weeks of age.

Figure 6.

Altered hepatic matrix deposition and inflammation 10 weeks after BM-Tx in Abcb4−/− mice. Photomicrographs of (H&E, A-C) and Masson Trichrome (D-F)-stained liver sections of sham-Tx control (A,D, Abcb4−/− mice, 16 weeks old), syngeneic BM-Tx mouse (B,E, Abcb4−/− mice, 16 weeks old, 10 weeks after syngeneic BM-Tx), and allogeneic BM-Tx mouse (C,F, Abcb4−/− mice, 16 weeks old, 10 weeks after allogeneic BM-Tx). BM-Tx enhanced piecemeal necrosis (B,C) but attenuated fibrotic staging (green staining). Well-defined borders of fibrotic septae (dashed white line) were abolished by piecemeal necrosis. Original magnification ×400, scale bar = 25 μm. Figures are representative images for all animals of each particular group.

In contrast to the increasing amounts of collagen, the hepatic α-SMA protein level did not change between weeks 4 and 16 (Fig. 1B). Type I collagen appeared largely without colocalization of α-SMA-positive cells within the liver parenchyma of 16-week-old Abcb4−/− mice (Fig. 2A,B).

Periductular Myofibroblasts in Proliferating Bile Ducts Originate from HSC.

Immunostaining revealed that approximately half of the proliferating bile ducts were accompanied by periductal CRP-2+ myofibroblasts (Fig. 2C,D). CRP-2 transcription was increased by 1.4-fold at the age of 8 weeks, whereas the quiescent-HSC-marker glial fibrillary acidic protein (GFAP) was not altered compared to wildtype animals (Fig. 2E).

CD34+ Fibrocytes Differentiate to a Desmin+ Matrix-Producing Cell Type.

Compared to wildtype BALB/c-mice, CD34 mRNA expression increased from 5.8-fold after 4 weeks to 14-fold after 16 weeks in Abcb4−/− mice (Fig. 3A). Desmin transcription was enhanced 5.2-fold after week 8 and 3.1-fold after week 16 in Abcb4−/− mice. Furthermore, CD34 and desmin gene expression were positively correlated with each other in BALB/c-Abcb4−/− mice (Fig. 3C; P < 0.001, r = 0.82). Enhanced hepatic CD34 transcription was accompanied by increased levels of stromal cell-derived factor-1 (SDF-1, up to 3.6-fold at week 8) and matrix metalloproteinase-9 (MMP-9, peaked at week 8, 17-fold). SDF-1 gene expression was correlated with CD34 (P = 0.007, r = 0.57), whereas MMP-9 appeared not to be associated with CD34 transcription. Fibrogenesis in Abcb4−/− mice was accompanied by increasing levels of MMP-13 (up to 11-fold at week 16) and tissue inhibitor of metalloproteinase-1 (TIMP-1) (up to 35-fold at week 16). Interestingly, MMP-2 was not altered in liver of Abcb4−/− mice.

Figure 3.

Transcriptional regulation of SDF-1, desmin, CD34, and MMPs. (A,B) Hepatic gene regulation of SDF-1, desmin, CD34, MMP-2, MMP-9, MMP-13, and TIMP-1. CD34 transcription increased by 13.9-fold with progressive fibrogenesis. SDF-1, desmin, and MMP-9 gene expression peaked with 3.6-, 5.2-, and 17-fold transcription levels, respectively, in 8-week-old Abcb4−/− mice. MMP-13 and TIMP-1 gene expression increased up to 11-fold and 34-fold, respectively. MMP-2 transcription was not altered. All data are normalized to r18S. Fold increase to wildtype animals is depicted (mean ± SD, n = 5). (C) The correlation of CD34 and desmin transcription in Abcb4−/− mice is demonstrated by scatterplot. Two-sided Spearman's rank correlation revealed a positive correlation of desmin and CD34 mRNA.

With the onset of inflammation and degeneration of ductular structures CD34+ cells initiate the occupation of portal fields. Progressive portal fibrosis was accompanied by the appearance of CD34+ cells whose distribution pattern clearly shaped the portal tracts (green staining, Fig. 4A,B). Desmin+ cells appeared at the border of altered portal fields and spread along sinusoids throughout parenchyma with progressive fibrosis (red staining, Fig. 4A-D). Desmin+ cells adjacent to portal fields revealed coexpression of CD34 (yellow arrows, Fig. 4A), whereas CD34+-phenotype disappeared in the parenchyma (Fig. 4A,B). Expression of collagen-type I prodomain (green staining) was frequently detected in desmin+ cells (red staining) at the border of portal fields as well as in almost all desmin+ cells within the fibrotic parenchyma (Fig. 4C,D).

Figure 4.

BM-derived CD34+ fibrocytes infiltrated portal tracts and differentiated into a desmin+ type I collagen expressing phenotype in Abcb4−/− mice. (A,B) Costaining of desmin (red) and CD34 (green). Although CD34+ cells were detected in portal tracts only, desmin+ cells appeared at both the edge of the portal tracts and throughout the parenchyma. The costaining in portal borders (yellow arrows) illustrated the differentiation of CD34+ progenitors into desmin+ cells (scale bar = 20 μm, original magnification ×400; B depicts a magnified part of A). (C,D) Coimmunofluorescence staining of type-I procollagen (green) and desmin (red) demonstrated the expression of type-I collagen from desmin+ cells within the parenchyma (scale bar = 10 μm, original magnification ×1,000; D depicts a magnified part of C). All mice (control and treatment) were 16 weeks of age. (E,F) In vitro coimmunofluorescence staining of type-I procollagen (green) and CD34 (red) on MACS-isolated fibrocytes on day 1 (E) and day 5 (F) after in vitro cultivation (scale bar = 5 μm, original magnification ×1,000). Cells were isolated from 16-week-old Abcb4−/− mice.

In Vitro Expression of Collagen Type I by Isolated CD34+ Fibrocytes.

In order to prove that CD34+ fibrocytes indeed express collagen type I, CD34+ cells were isolated from Abcb4−/− liver cell suspension by MACS. Prior to the enrichment of CD34+ cells, CD31+ endothelial cells were depleted by MACS. Purity of MACS-enriched fibrocytes was 88% determined by fluorescence-activated cell sorting (FACS). Isolated CD34+/CD31− fibrocytes were cultured on collagen type I-coated glass coverslips. On day 1 and day 5 after plating CD34+/pro-collagen type I-expressing fibrocytes were identified immunohistochemically (Fig. 4E,F). During the first 5 days of cultivation, CD34+ fibrocytes developed a spindle-like fibroblast phenotype (Fig. 4F).

CD34+ Fibrocytes Originate from BM Stem Cells.

To distinguish between BM-derived CD34+ fibrocytes from resident liver cells, BM of GFP+ transgenic donors was transplanted into Abcb4−/− mice. Hepatic GFP+ cells were identified by fluorescence microscopy. Immunostaining demonstrated expression of GFP in 89.6 ± 2.5% of CD34+ cells (GFP green; CD34 red; costaining yellow; Fig. 5A) and 77.6 ± 7.4% of desmin+ cells (desmin red, Fig. 5B). Coexpression of GFP and desmin was observed adjacent to portal tracts, whereas desmin+ fibrocytes in parenchyma happened to be GFP. We thus hypothesize that CD34+/desmin cells migrate by way of portal fields into hepatic lobules and then develop a desmin+ phenotype within the fibrotic parenchyma.

Figure 5.

Immunostaining and FACS analysis proved BM engraftment of transplanted cells. (A) Coimmunostaining of GFP (green) and CD34 (red) on liver slices of 16-week-old Abcb4−/− mice, 10 weeks after BM-Tx, proved the hematopoietic origin of CD34+ cells in portal tracts (yellow arrows: CD34+/GFP+ fibrocytes, scale bar = 10 μm, original magnification ×1,000). (B) Coimmunofluorescence staining of GFP (green) and desmin (red) demonstrated the BM origin of mature, desmin+ fibrocytes (yellow arrows: desmin+/GFP+ mature fibrocytes; scale bar = 10 μm, original magnification ×1,000). (C) Coimmunofluorescence staining of GFP (green) and α-SMA (red) demonstrated that periductular myofibroblasts were not derived from BM (red arrows: periductular α-SMA+/GFP myofibroblast, scale bar = 10 μm, original magnification ×1,000). Representative micrographs are shown. (D) Expression of epithelial-marker CK-7 (green) and EMT-marker S100A4 (red) in proliferating bile ducts. Coexpression of both CK-7 and S100A4 identifies activated cholangiocytes undergoing EMT (scale bar = 10 μm, original magnification ×1,000). Efficiency of BM-Tx was analyzed by flow cytometry of peripheral blood cells in (E) sham-Tx mice and (F) GFP+-BM Tx animals 10 weeks after Tx. Histograms illustrate that ≈70% of peripheral leukocytes showed GFP-fluorescence, indicating donor origin.

In contrast, no GFP+ cells were present in livers of sham-operated mice (data not shown). The effectiveness of GFP+-BM engraftment was evaluated by flow cytometry. Approximately 70% of the recipient leukocytes were replaced by donor cells 10 weeks after Tx, indicating a considerable exchange of the original stem cell population (Fig. 5E,F).

EMT Contributes to Fibrogenic Cell Pools.

Cholangiocytes within proliferating bile ducts demonstrated coexpression of endothelial-marker CK-7 and EMT-marker S100A4 (yellow arrow, Fig. 5D). Cells solely stained for S100A4 represent matrix-producing myofibroblasts adjacent to proliferative bile ducts. Note that only the cholangiocyte epithelial monolayer is stained for CK-7.

BM Transplantation Ameliorates Liver Staging, Serum AST, and Serum AP Compared to Sham-Tx Animals.

The effects of BM-Tx on liver fibrosis, inflammatory grade, and functional serum parameters were analyzed. Ten weeks after BM-Tx, livers from GFP+ transplanted Abcb4−/−− mice displayed a decreased fibrosis staging compared to nontransplanted Abcb4−/−− controls (F = 1.6 ± 0.35 versus F = 2.0 ± 0.0, P = 0.015, κ = 0.74). This effect came along with improved AST (255 ± 64 U/L versus 686 ± 384 U/L, P = 0.003) and AP serum levels (348 ± 64 U/L versus 494 ± 1694, P = 0.033), whereas ALT only gradually decreased (294 ± 62 U/L versus 419 ± 176 U/L, P>0.1) (Fig. 6; Table 2). No significant differences were noticed among allogeneic and syngeneic BM-Tx (Abcb4−/−Abcb4−/−).

Table 2. Serum Liver Function Parameters, Staging, and Grading of Fibroinflammatory State
 AST (U/L)ALT (U/L)AP (U/L)StagingHyp/μg/gGrading
  1. Total collagen determination in sham-controls and BM-Tx animals 10 weeks after transplantation.

Sham (n = 8)686 ± 384419 ± 176494 ± 1692 ± 0.0456 ± 1093.29 ± 0.95
 433 ± 264393 ± 192403 ± 1181.8 ± 0.27462 ± 705.8 ± 0.98
Syngeneic Tx (n = 8)PSh = 0.10PSh = 0.57PSh = 0.046PSh = 0.040PSh = 0.89PSh = 0.004
 255 ± 64294 ± 62348 ± 641.63 ± 0.35506 ± 1484.88 ± 2.17
Allogeneic Tx (n = 8)PSh = 0.003PSh = 0.18PSh = 0.033PSh = 0.015PSh = 0.23PSh = 0.26

Neither hepatic collagen accumulation assessed by hydroxyproline level nor inflammatory grading differed between syngeneic and allogeneic BM-Tx mice.

Graft-Versus-Host Disease-Related Cytokines Enhanced After BM-Tx.

Transcriptional levels of IL-1β and TNF-α in GFP+ transplanted Abcb4−/−− mice were enhanced 3.7 ± 0.6-fold and 3.2 ± 0.3-fold, respectively, compared to sham-Tx-Abcb4−/− mice (both P = 0.008) (Fig. 7).

Figure 7.

Enhanced transcriptional levels of IL-1β and TNF-α indicate graft-versus-host disease after BM-Tx. Hepatic gene expression of IL-1β and TNF-α increased by 3.7-fold and 3.2-fold, respectively, 10 weeks after BM-Tx from GFP+ mice in Abcb4−/−− mice. All data were normalized to r18S. Fold increase versus sham-Tx Abcb4−/− animals of the same age is presented (mean ± SD, n = 8).


Abcb4−/− mice develop biliary fibrosis shortly after birth as a result of leakage of potentially toxic bile acids into the periductular area.7 Thereby, bile ducts demonstrate age-dependent enlargement, progression of inflammation, and proliferation. Periductular accumulation of extracellular matrix is followed by periportal and bridging fibrosis in liver. Histological findings in BALB/c-Abcb4−/− mice are consistent with our previously published reports and other recently published data for FVB/-Abcb4tm1Bor mice.7, 17 Moreover, liver fibrosis in BALB/c-Abcb4−/− mice is characterized by diffuse periportal collagen aggregation as well as intralobular collagen deposition along the liver sinusoids.17

The pathophysiological rationale for the development of liver fibrosis has been established by Popov et al.7: They described transiently enhanced transcription of α-SMA in 4-week-old Abcb4−/− mice. Therefore, α-SMA+ myofibroblasts seemed to be the most important fibrogenic cell population in the early stages of the development of biliary fibrosis.

In the present study approximately half of the proliferating bile ducts were accompanied by α-SMA+ myofibroblasts derived from HSC. Thus, α-SMA+ myofibroblasts are presumably the source of periductular fibrogenesis. However, we did not detect α-SMA+ myofibroblasts adjacent to periportal and septal liver fibrosis in Abcb4−/− mice (Fig. 2). Although hepatic α-SMA protein levels were not altered between week 4 and week 16 (Fig. 1B), the amount of type I collagen and HYP significantly increased over time (Figs. 1A, 2A,B). In the liver parenchyma of 16-week-old mice, i.e., those animals with advanced hepatic fibrosis, colocalization of type I collagen with α-SMA-positive cells was absent. Addressing these data we doubt that α-SMA+ myofibroblasts are the only source for fibrogenesis in sclerosing cholangitis.

We therefore set off to identify other key mediators and potential cellular sources for advanced liver fibrosis. Herein we demonstrate that progressive fibrosis in Abcb4−/− mice is accompanied by infiltration of BM-derived CD34+ fibrocytes into hepatic lobules. Immunohistologic staining revealed a subsequent transition of CD34+ fibrocytes into CD34/desmin+ pro-type I collagen-expressing cells at the border of portal tracts. Procollagen type I expression was allocated to almost all desmin+ parenchymal cells. Furthermore, CD34+ fibrocytes were isolated from Abcb4−/− mouse liver and collagen type I expression was proven in vitro. Thus, BM-derived fibrocytes considerably account for advanced fibrogenesis in mature Abcb4−/− mice.

SDF-1 and MMP-9 are expressed by liver bile duct epithelium and mediate directional migration of CD34+ stem cells to the liver.26 The presented data serve to emphasize that the increase in SDF-1 and MMP-9 levels are not only related to stem cell recruitment, but may rather represent a basic pathological principle of liver fibrogenesis in cholangitis. Enhanced expression of MMP-9, MMP-13, and TIMP-1 (up to 35-fold at week 16) fits nicely with other models of hepatic fibrosis. Gene expression of MMP-2 is not regulated in Abcb4−/− mice and may reflect a specific pathogenesis of fibrosis in these mice.27

Since the 1990s, CD34+ fibrocytes have been associated with a variety of physiologic and pathologic conditions including hepatic fibrosis.8, 10 Recently it has been demonstrated that fibrocytes of BM origin infiltrate portal tracts and hepatic parenchyma in a murine model of bile duct ligation (BDL).15 Herein, BM-derived fibrocytes contributed merely 5%–10% of type I collagen-expressing cells in BDL-injured liver irrespective of desmin+ fibrocytes.15 In a murine model of toxic hepatic fibrosis 33% of all HSC expressed characteristic BM-marker molecules, such as CD40 and CCR5.28 This once again suggests a causal relationship of fibrocytes and fibrogenesis, i.e., fibrocytes may represent progenitors of fibrogenesis. After liver transplantation with subsequent transplant fibrosis, fibrocytes of recipient origin were found within fibrotic areas.14

Russo et al.16 demonstrated in both CCl4 and TAA models of liver fibrosis in mice that approximately 70% of α-SMA-positive myofibroblasts and a similar proportion of HSC originate from the BM. The authors concluded that BM is of minor relevance for hepatocyte regeneration, but constitutes an important source for functional HSC and myofibroblasts in toxic liver injury. Thus, BM contributes significantly to fibrogenesis.16 After allogeneic-Tx we found GFP reporter protein in cells functionally contributing to hepatic fibrosis. Of note, in our model of Abcb4−/− mice CD34+ and desmin+ fibrocytes originated from BM, whereas hepatocytes did not.

Di Bonzo et al.29 transplanted human mesenchymal stem cells into immunodeficient NOD/SCID mice. Induction of chronic liver fibrosis by way of CCl4 resulted in 7% liver engraftment from human BM cells. Half of these cells showed a myofibroblast-like phenotype, whereas the other half were described as “small cells with low nucleus/cytoplasm ratio.” The latter presented a phenotype comparable with CD34+ fibrocytes in our mouse model of Abcb4−/− mice. Herein, the extent of liver fibrosis was positively associated with the infiltration of BM-derived CD34+ fibrocytes, subsequently differentiating into a CD34/desmin+/α-SMA fibrocyte phenotype (Fig. 4).

Taken together, the BM functionally contributed to fibrogenesis in Abcb4−/− mice irrespective of the presence or absence of α-SMA, thereby supporting data from toxically induced liver fibrosis triggered by myofibroblasts.

Just recently in CCl4- and bile duct ligation-induced hepatic fibrogenesis in mice the impact of BM-derived cells on fibrogenesis was found “negligible.”30 Of note, experimental setups and analyses differed considerably. This may once again serve to emphasize the importance of the particular model (i.e., the etiology of tissue injury and timeframe) and the method of analysis (i.e., applied surrogate parameters reflecting fibrosis or fibrogenesis, respectively). The existing variability in models and methods may facilitate conflicting results, thereby affecting our current understanding of the role of BM-derived cells with regard to fibrosis and fibrogenesis.

EMT may contribute to the pathogenesis of biliary fibrosis, in particular the appearance of α-SMA+ periductular myofibroblasts.31, 32 We demonstrated that peribiliary α-SMA+ myofibroblasts in Abcb4−/−− mice stem from HSC (Fig. 2). Moreover, cholangiocytes expressing CK-7 as well as S100A4, a marker for EMT by immunohistological costainings, were visualized (Fig. 5D).31, 32 Thus, EMT is present in the areas of proliferating bile ducts of Abcb4−/− mice. Nevertheless, CK-7 is restricted to the epithelial cholangiocyte monolayer, lacking signs of cholangiocyte proliferation and transdifferentiation.

In order to circumvent both any BM-Tx-related side-effects and a nonsignificant reduction in liver fibrosis we focused on a long-term follow-up, i.e., 10 weeks after transplantation.

Although BM-Tx (i.e., transplantation of GFP+/Abcb4+ BM) conducted within this study improved liver function and staging compared with sham transplantation, no significant differences were noticed among allogeneic and syngeneic transplantation (Table 2). This suggests that irrespective of the genotype of transplanted BM, changes in immunity (e.g., graft-versus-host reaction) may alter fibrosis. The significant increase of IL-1β and TNF-α, usually considered the most important cytokine mediators of graft-versus-host disease,33 support this notion (Fig. 7). Albeit some evidence may encourage this train of thought, it is important to stress that on the basis of our experiments we are unable to prove this. Thus, it remains rather speculative if graft-versus-host disease is indeed associated with the observed effects.

A growing number of studies investigated the therapeutic capabilities of BM-Tx in chronic liver disease.34–36 However, a critical review of the literature suggests that engraftment and hepatocyte differentiation of BM-derived cells remains a controversial issue.37

In conclusion, the present study is the first to identify that both BM-derived fibrocytes and HSC are involved in biliary fibrogenesis in Abcb4−/− mice. BM-Tx may well serve as a therapeutic option for the treatment of sclerosing cholangitis or liver fibrosis. Our data suggest that changes in immunity subsequent to BM-Tx may alter fibrosis.


The authors thank Dr. M. Heil (Max-Planck-Institute, Bad Nauheim, Germany) for BALB/c-GFP+ mice, R. Salguero-Palacios and A. Tschuschner for excellent technical assistance, and Dr. Walbott and Dr. Lugert from Strahlenzentrum Justus-Liebig-University Gieβen for irradiation experiments.