Human mesenchymal stem cells are recruited to injured liver in a β1-integrin and CD44 dependent manner


  • Victoria Aldridge,

    Corresponding author
    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    • NIHR Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, United Kingdom===

    Search for more papers by this author
    • fax: (44)-214158701

  • Abhilok Garg,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Nicholas Davies,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • David C. Bartlett,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Janine Youster,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Heather Beard,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Dean P. Kavanagh,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Neena Kalia,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Jon Frampton,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Patricia F. Lalor,

    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Philip N. Newsome

    Corresponding author
    1. National Institute for Health Research Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, UK
    • NIHR Biomedical Research Unit and Centre for Liver Research, Institute of Biomedical Research, University of Birmingham, Birmingham B15 2TT, United Kingdom===

    Search for more papers by this author
    • fax: (44)-214158701

  • Potential conflict of interest: Nothing to report.


Human bone marrow mesenchymal stem cells (hMSCs) have shown benefit in clinical trials of patients with liver disease. Efficient delivery of cells to target organs is critical to improving their effectiveness. This requires an understanding of the mechanisms governing cellular engraftment into the liver. Binding of hMSCs to normal/injured liver tissue, purified extracellular matrices, and human hepatic sinusoidal endothelial cells (HSECs) were quantified in static and flow conditions. To define the mechanisms underpinning hMSC interactions, neutralizing adhesion molecule antibodies were used. Fluorescently labelled hMSCs were infused intraportally into CCl4–injured mice with and without neutralizing antibodies. hMSCs expressed high levels of CD29/β1-integrin and CD44. Using liver tissue binding assays, hMSC adhesion was greatest in diseased human liver versus normal liver (32.2 cells/field versus 20.5 cells/field [P = 0.048]). Neutralizing antibodies against CD29 and CD44 reduced hMSC binding to diseased liver by 34% and 35%, respectively (P = 0.05). hMSCs rolled at 528 μm/second on HSECs in flow assays. This rolling was abolished by CD29 blockade on hMSCs and vascular cell adhesion molecule-1 (VCAM-1) blockade on HSECs. Firm adhesion to HSECs was reduced by CD29 (55% [P = 0.002]) and CD44 (51% [P = 0.04]) blockade. Neutralizing antibodies to CD29 and CD44 reduced hepatic engraftment of hMSCs in murine liver from 4.45 cells/field to 2.88 cells/field (P = 0.025) and 2.35 cells/field (P = 0.03), respectively. hMSCs expressed modest levels of chemokine receptors including CCR4, CCR5, and CXCR3, but these made little contribution to hMSC adhesion in this setting. Conclusion: hMSCs bind preferentially to injured liver. Rolling of hMSCs is regulated by CD29/VCAM-1, whereas CD29/CD44 interactions with VCAM-1, fibronectin, and hyaluronan on HSECs determine firm adhesion both in vitro and in vivo as demonstrated using a murine model of liver injury. (HEPATOLOGY 2012;56:1063–1073)

Human mesenchymal stem cells (hMSCs) are multipotent cells originating principally from bone marrow that can differentiate into adipocytes, osteoblasts, and chondrocytes and have the capacity for self-renewal.1 hMSCs have been identified within the liver, where various actions of these cells have been reported.2 These include transdifferentiation into hepatocytes, stimulation of endogenous hepatocyte proliferation, and suppression of immunomediated rejection.3-8 hMSC contribution to liver fibrogenesis has been reported, although this is not a universal finding.9-11 The delivery of hMSCs to injured organs is important in the mediation of their effect, hence the manipulation of hMSC numbers in the injured liver is an important therapeutic goal that requires an understanding of the mechanisms regulating their egress into the liver.12 It is unclear whether the leukocyte paradigm of vascular adhesion applies to hMSC recruitment into the liver.13 However, hMSCs roll on human umbilical vein endothelial cells in a P-selectin–dependent manner and use VLA-4 (α4β1) and vascular cell adhesion molecule-1 (VCAM-1) for firm adhesion.14 We reported VCAM-1 expression in portal tracts in normal liver, and this expression is increased in disease and expanded to hepatic sinusoidal endothelial cells (HSECs).15 Expression of other candidate ligands that may be required for hMSC recruitment such as fibronectin and hyaluronan have been similarly reported.15, 16 Indeed, the hyaluronate receptor CD44 is involved in the recruitment of hemopoietic and lymphocyte populations to the liver,15-18 and hyaluronan is specifically expressed on HSECs.16 Additionally, hMSCs are known to express a restricted repertoire of chemokine receptors and can migrate to specific chemokines.19, 20 Taken together, these studies implicate chemokine receptors, integrins, and CD44 in controlling adhesion of hMSCs to liver. However, to date no study has specifically addressed the mechanisms involved in mediating adhesion of hMSCs to injured liver tissue.

We demonstrate that hMSCs are capable of binding matrix and endothelial adhesive ligands, whose expression is increased during liver injury. hMSCs undergo a rapid capture interaction with HSECs under flow, which is mediated by CD29/VCAM-1. Furthermore, both CD29 and CD44 support firm adhesion to human and mouse liver endothelium. Interestingly, G-protein–coupled receptors appear to play a minimal role in hMSC adhesion to liver. We have identified key adhesive interactions that may be targeted to modify delivery and retention of hMSCs within the liver for therapeutic applications.21


ALD, alcoholic liver disease; CFSE, carboxyfluorescein diacetate succinimidyl ester; hMSC, human mesenchymal stem cells; HSEC, hepatic sinusoidal endothelial cell; HUVEC, human umbilical vein endothelial cell; IFN-γ, interferon-γ; IgG, immunoglobulin G; MFI, mean fluorescent intensity; PBC, primary biliary cirrhosis; PTX, pertussis toxin; TEM, transendothelial migration; TNFα, tumor necrosis factor-α; VCAM-1, vascular cell adhesion molecule-1.

Materials and Methods

Adhesion of hMSCs to Purified Recombinant Proteins and HSECs.

An Ibidi μ-slide VI (Ibidi) was coated with protein (VCAM-1 10 μg/mL [R&D Systems], fibronectin 50 μg/mL [Sigma], or hyaluronan 2.5 mg/mL [Sigma]) for 2 hours at 37°C. Nonspecific protein-binding sites were blocked using 0.1% bovine serum albumin (BSA) for 30 minutes. For HSEC experiments, Ibidi μ-slides were coated with rat tail collagen for 30 minutes prior to seeding with HSECs. HSECs were treated for 24 hours with 10 ng/mL tumor necrosis factor-α (TNFα)/interferon-γ (IFNγ) (Peprotech) prior to assay. When appropriate, HSECs were treated with 30 U/mL hyaluronidase (Sigma) for 15 minutes prior to the assay. hMSCs were trypsinized, resuspended at 1 × 106 cells/mL in serum-free media and perfused over protein or endothelium at 0.5 dynes/cm2 to mimic physiological shear stress. hMSCs and adhesive substrate interactions were recorded for 2 minutes to permit calculation of rolling velocities. Binding was maximized by stopping flow for 5 minutes, and the number of cells in 10 fields of view was recorded. Flow was then restarted with wash buffer for 5 minutes, and cells remaining adherent were similarly recorded. Data are expressed as the percentage of total adherent cells binding in the presence of blocking antibodies (CD29, CD44, VCAM-1, or appropriate isotype controls) (Supporting Table 1).

MSC Engraftment in Chronically Injured Murine Liver.

6- to 10-week-old C57BL/6 mice received twice-weekly intraperitoneal CCl4 (0.25 mL/kg [Sigma]) for 4 weeks to induce chronic liver injury (evaluated with picrosirius red stain). A total of 1 × 106 carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled hMSCs treated with neutralizing antibody/immunoglobulin G (IgG) control (Supporting Table 1) were introduced into anesthetized mice via the portal vein. Cells were allowed to engraft for 15 minutes before sacrificing and liver freezing in liquid nitrogen. Livers were cut into 5-μm sections, and the mean number of adherent fluorescent cells was determined from 10 random fields of view (magnification ×200; Axioskop 40 Zeiss, UK).

Additional details are provided in the Supporting Materials and Methods.


hMSCs Express Characteristic Markers and Differentiate Appropriately.

hMSCs expressed the characteristic markers CD90, CD105, and CD73 and were negative for the hematopoietic markers CD34 and CD45, the B cell marker CD20, and the monocyte/macrophage marker CD14 (Fig. 1A). Tri-lineage differentiation into fat, cartilage, and bone demonstrated hMSC functionality (Fig. 1B,C).

Figure 1.

hMSCs expressed characteristic markers and differentiated appropriately. hMSC surface receptor expression was determined by flow cytometry. (A) Receptor expression dot plots are shown. NEG = a mix of antibodies CD34, CD45, CD20, and CD14. (B) Oil red O and collagen II staining demonstrate adipogenic and chondrogenic differentiation, respectively. (C) Calcium production suggests osteogenic differentiation.

Impact of hMSC Passage on Expression of Adhesion and Chemokine Receptors Necessary for Binding and Migration.

Flow cytometry was used to analyze the expression of candidate receptors that may mediate adhesion of hMSCs to the liver at hMSC passages 4 and 6. Differentiation ability at these passages was unaltered; however, at passage 6, the growth rate slowed and a flattened morphology developed (data not shown). hMSCs expressed CD29/β1-integrin, CD49a/α1-integrin, CD49c/α3-integrin, CD49e/α5-integrin, αVβ5-integrin, and CD44 at high levels at both passages (Fig. 2A,B). CD49b/α2-integrin and CD49d/α4-integrin were expressed on a moderate proportion of cells at passage 4 (39% and 47%, respectively). αVβ3-integrin was expressed by a smaller percentage of hMSCs and other integrins, including CD49f/α6-integrin; CD18/β2-integrin and CD61/β3-integrin were undetectable (data not shown). At passage 6, the percentage expression of both CD49b and CD49d was significantly increased to 85% (P = 0.039) and 68% (P = 0.04), respectively (Fig. 2A). The mean fluorescent intensity (MFI) for CD49b, CD49c, and αVβ5 were significantly increased at passage 6 (Fig. 2B). CD49b increased from 17 to 44 (P = 0.048), CD49c increased from 59 to 269 (P = 0.013), and αVβ5 increased from 39 to 120 (P = 0.041). hMSCs expressed a restricted set of surface chemokine receptors, with the most highly expressed being CCR4 (28.5%), CCR5 (19.33%), and CXCR3 (17.20%). Lower levels of CCR6 (4.84%), CCR9 (5.38%), CCR10 (5.31%), and CXCR1 (6.61%) were seen (Fig. 2C,D). MFI values for surface expression of all receptors were low (data not shown), but intracellular stores of CCR4 (96%), CCR5 (68%), CXCR3 (93%), CCR6 (19%), CCR9 (13%), CCR10 (14%) and CXCR1 (22%) were found (Fig. 2D). In modified Boyden chamber assays hMSCs migrated to CCR4 ligands CCL22 and CCL17 and the CXCR3 ligand CXCL11 (Fig. 2E). No significant migration was seen to any of the CCR5 ligands tested including CCL8 or CCL4. Since increases in certain adhesion molecules may be associated with a more committed cell population, we subsequently only used hMSCs at passage 4.

Figure 2.

Impact of hMSC passage on expression of adhesion and chemokine receptors necessary for binding and migration to the liver. hMSCs from different donors were stained with monoclonal antibodies directed against specific integrin subunits/heterodimers, CD44 and chemokine receptors and analyzed by flow cytometry. (A, B) hMSC integrin and CD44 expression was determined at passages 4 and 6. The % integrin expression is shown in (A); the MFI is shown in (B). (C, D) Surface (C) and intracellular (D) chemokine receptor percentage expression is also shown. Symbols indicate individual experiments; bars represent the mean. (E) Migration of hMSCs was assessed in the modified Boyden chamber assay and is shown as chemotactic index, which is the ratio of mean migrated cells versus control. Data are expressed as the mean of three independent experiments. *P < 0.05.

hMSCs Preferentially Bind Diseased Liver.

Modified Stamper Woodruff assays confirmed that hMSCs are capable of binding normal and injured human liver tissue. Increased numbers of hMSCs were found to bind to alcoholic liver disease (ALD) and primary biliary cirrhosis (PBC) compared with normal liver (Fig. 3A) (mean cell numbers: normal liver, 20.5 ± 7.7; ALD, 35 ± 9.6 [P = 0.05]; PBC, 30 ± 9.6 [P = 0.04]). Of note, hMSCs bound avidly to the matrix-rich fibrotic septa in diseased liver sections (Fig. 3B, middle panel, arrows) and to CD31+ endothelial structures (Fig. 3B, right panel, arrows). We used a flow-based assay to compare hMSC binding with unstimulated HSECs versus HSECs stimulated with TNFα/IFNγ for 24 hours (to replicate inflammatory conditions in injured livers). Approximately twice as many cells bound to stimulated HSECs (Fig. 3C; 11.2 ± 1.04 versus 5.9 ± 1.06; P = 0.025) compared with unstimulated HSECs.

Figure 3.

hMSCs preferentially bind to diseased liver. Static adhesion assays with hMSCs were performed using sections of liver tissue (normal liver [NL], PBC, ALD) to determine hMSC-liver binding ability. (A) Mean cell numbers binding to normal liver and the diseased liver sections are shown. (B) Photomicrographs of whole liver sections show enhanced binding to matrix-rich septa (middle panel, arrow) and CD31+ endothelial structures (far right panel, arrows). (C) hMSCs were perfused over untreated or TNFα/IFNγ-treated HSECs in a modified flow assay. The mean number of cells binding is shown. Data are expressed as the mean of three independent experiments. *P < 0.05.

Chemokine-Independent hMSC Adhesion to Liver Tissue Is Mediated by VCAM-1, CD44, and CD29.

Neutralizing antibodies to CD29, CD44, and VCAM-1 or cells treated with pertussis toxin (PTX) were used in a modified Stamper-Woodruff assay with normal and diseased human liver. CD29 blockade reduced binding to injured but not normal liver (Fig. 4A). Specifically, CD29 block significantly reduced the binding of hMSCs to the sinusoids of ALD liver (20% [P = 0.04]), and both the sinusoids and portal tracts of PBC liver compared with IgG (41% [P = 0.01] and 41% [P = 0.01], respectively) (Fig. 4B,C). Blockade of the CD29 ligand VCAM-1 significantly reduced binding to portal tracts in normal liver compared with IgG (21% [P = 0.002]) (Fig. 4A). VCAM-1 blockade on injured liver sections resulted in a greater reduction in hMSCs binding to sinusoids (PBC, 60% [P = 0.008]; ALD, 49% [P = 0.01]; as well as portal tracts of ALD, 27% [P = 0.04]) (Fig. 4B,C). CD44 blockade resulted in a significant reduction in hMSCs binding to sinusoids of both normal liver (28% [P = 0.04]) and injured liver (PBC, 35% [P = 0.01]) compared with IgG control (Fig. 4A,B). There was no reduction in binding to normal or diseased liver tissue sections when hMSCs were pretreated with PTX (Fig. 4).

Figure 4.

hMSC adhesion to liver tissue is mediated by VCAM-1, CD44, and CD29. Static adhesion assays were performed using sections of liver tissue (normal [A]; PBC [B]; ALD [C]) and hMSCs. hMSCs, liver tissue, or both were treated with neutralizing antibodies before use in the assay. Data represent total adhesion as a percentage of that seen in the absence of blocking antibody (IgG control) and are expressed as the mean of four independent experiments. *P < 0.05 versus IgG control.

VCAM-1 Mediates Rolling of hMSCs Under Flow and Along With Hyaluronan and Fibronectin Mediates Firm Adhesion and Adoption of a Pro-Motility state.

hMSC binding to immobilized proteins was studied under conditions replicating liver sinusoidal blood flow (0.5 dynes/cm2). Experimental videos revealed fast-rolling interactions between phase-bright hMSCs and immobilized VCAM-1 at a mean velocity of 448 μm/second ± 17.06 μm/second, which was not observed on fibronectin or hyaluronan (Fig. 5A and Supporting Video 1). Perfusing CD29-neutralized hMSCs over VCAM-1 abolished rolling confirming that the interactions were adhesion-dependent (Fig. 5B and Supporting Video 2). Next, we performed modified-flow experiments to recreate the combination of passive and active deceleration suggested for hMSCs.13 Briefly, cells were flowed over matrix and allowed to rest for 5 minutes, followed by reapplication of physiological shear. The percentage of arrested cells was then quantified. Whereas few cells remained adherent to BSA control using this methodology, the majority of hMSCs that initially bound to fibronectin remained firmly adhered (92%) (Fig. 6A). Similar results were found with hyaluronan (77% remained bound) and VCAM-1 (63% remained bound) (Fig. 6A). hMSC binding to VCAM-1 was reduced by 66% (P = 0.0001) when CD29 was neutralized (Fig. 6B). Blocking CD29 or CD44 on hMSCs perfused over fibronectin reduced firm adhesion by 28% (P = 0.04) and 34% (P = 0.02), respectively, compared with the IgG control (Fig. 6B). CD44 blockade reduced hyaluronan binding by 43% (P = 0.009) (Fig. 6B). Notably, a large proportion of the cells that remained firmly adhered to immobilized proteins (but not BSA) became activated and shape-changed upon the substrate. This effect was most marked on fibronectin (78% of adherent cells) and hyaluronan (69%) (Fig. 6A,C). These observations suggest the cells become activated on these extracellular matrix components and may develop a more motile morphology.

Figure 5.

VCAM-1 supports capture of hMSCs from flow mediated by CD29. hMSCs were perfused over VCAM-1, fibronectin, and hyaluronan in a modified flow assay. (A) Video stills of hMSC-HSEC interactions show rolling on VCAM-1 but not fibronectin/hyaluronan. (B) The flow over VCAM-1 was repeated, incorporating a CD29 neutralizing antibody that ablated rolling.

Figure 6.

hMSC adhesion to ligands up-regulated in liver disease is mediated by CD29 and CD44. hMSC–extracellular matrix interactions were studied in a modified flow adhesion assay. (A) hMSC firm adhesion data are shown as % adhesion, and shape change data are expressed as % adherent cells undergoing shape change. (B) Firm adhesion of hMSCs to VCAM-1, fibronectin, and hyaluronan in the presence of neutralizing antibodies to CD29 and CD44 is shown. (C) Representative images from flow assay videos show the morphology of the hMSCs when adhered to different ligands. Data in (B) are expressed as % IgG control and are the mean of three independent experiments. *P < 0.05.

hMSCs Undergo Fast Rolling on HSECs.

hMSC binding to receptors expressed in a physiological context on HSECs was studied. hMSCs demonstrated fast-rolling interactions on HSECs (Supporting Video 3 and Fig. 7A) at a mean velocity of 528 μm/second ± 46.69 μm/second, which were again abolished by neutralizing CD29 on hMSCs (Supporting Video 4) or blocking VCAM-1 on HSECs (Fig. 7A). CD44 neutralization or pretreatment of HSECs with hyaluronidase to disrupt hyaluronan or other glycosaminoglycans did not affect rolling (Fig. 7A,B). Moreover, hMSC rolling velocity was unaffected by either CD44 or hyaluronidase treatment (Fig. 7B).

Figure 7.

hMSCs fast-roll on HSECs which is mediated by CD29/VCAM-1 and adhere which is mediated by both CD29 and CD44. hMSC-HSEC interactions were studied in a modified flow assay. (A) Video stills show the effects of function-blocking antibodies on hMSC rolling. (B) The effects of blocking antibodies on hMSC velocity. (C) Firm adhesion of hMSCs to HSECs following blockade of CD29 on hMSCs, VCAM-1/hyaluronan on HSECs, and CD44 on both, as well as a combination blockade with all antibodies/treatments. Data are expressed as % adhesion and are the mean of four independent experiments. *P < 0.05.

hMSC Recruitment to HSECs Is Essentially Mediated by CD44 and CD29.

CD29 blockade on hMSCs reduced firm adhesion to HSECs by 55% (P = 0.01) (Fig. 7C). CD44 block on both hMSCs and HSECs reduced adhesion by 51% (P = 0.001) (Fig. 7C). VCAM-1 block on HSECs reduced hMSC binding by 25% (P = 0.02), and hyaluronidase treatment had similar inhibitory effects (33% [P = 0.0001]) (Fig. 7C), demonstrating the importance of these HSEC proteins. A combination block (all antibodies/treatments, except PTX) did not reduce adhesion further than individual blockade, but reduced adhesion compared with the control (40% [P = 0.0001]) (Fig. 7C). Pretreatment of hMSCs with PTX before perfusion did not reduce firm adhesion (Fig. 7C).

Blockade of CD29 and CD44 Reduces hMSC Engraftment In Vivo.

To confirm our data in an in vivo setting, we intraportally injected CFSE-labelled hMSCs into chronically injured (CCl4) mice, and the number of cells remaining in the explanted livers was quantified. CCl4-induced injury was confirmed histologically using picrosirius red staining and was equal in all animals (Fig. 8A). In mice injected with isotype antibody–treated hMSCs, a mean of 4.45 ± 0.85 cells were seen per field of view (Fig. 8B). This was reduced to 2.88 ± 0.64 (P = 0.025) with CD29 blockade and 2.35 ± 0.63 (P = 0.03) with CD44 blockade (Fig. 8B). A combination block reduced the number of cells per field to 1.86 ± 0.32 (P = 0.03), although this was not greater than individual CD44 or CD29 block (Fig. 8B). Analysis of hMSC morphology within the injured livers showed that some hMSCs exhibited shape change in vivo (arrow on image Fig. 8C). Fluorescent immunostaining of the liver sections demonstrated that the recruited control hMSCs were closely surrounded by fibronectin and VCAM-1 and were located near CD31+ vessels (Fig. 8D). hMSCs are heterogenous in size, ranging from 5 to 50 μm in suspension culture and in liver (Supporting Fig. 1A,B). This large size may lead to physical entrapment rather than engagement of adhesion receptors within the liver. However, there was no difference in cell size, with control cells averaging 25μm ± 2.5 μm, CD29 blockade averaging 25 μm ± 2.3 μm, and CD44 blockade averaging 24 μm ± 1.9 μm (Supporting Fig. 1C). Cellular transplantation can result in local ischemia, which may affect hMSC adhesion, although this usually takes 1-4 hours to develop22 (longer than our experiments). Nevertheless, we studied hepatic hMSC adhesion in a murine model of ischemia-reperfusion injury. We found that ischemia-reperfusion injury did not result in an increase in hepatic hMSC adhesion (Supporting Fig. 1D), suggesting that our adhesion data are minimally affected by ischemic changes to the liver.

Figure 8.

Blockade of CD29, CD44, or a combination reduces hMSC engraftment in vivo. (A) CCl4-induced injury was confirmed by picosirius red staining of collagen using standard methodology. A representative field is shown from an injured and an uninjured animal (magnification ×10). (B) The number of CFSE-labeled hMSCs in injured murine liver following portal vein injection was determined after treatment with function-blocking antibodies to CD29, CD44, a combination blockade, or the relevant IgG control. The mean number of cells binding in each case is shown. (C) Representative image showing CFSE-hMSCs in murine liver. hMSCs seem to undergo shape change in vivo (arrow). (D) Engrafted hMSCs (green) can be seen surrounded by fibronectin (red) and VCAM-1 (red) and are often located near CD31+ (red) vessels as shown by immunostaining of IgG-treated CFSE-hMSC murine liver sections. Data are expressed as the mean of four independent experiments. *P < 0.05.


We demonstrate that hMSCs adhere to liver tissue/endothelium and bind in greater numbers in diseased conditions. Our functional blockade experiments suggest that these adhesive events are largely regulated by CD29 and CD44 in vitro and in vivo.

The integrin/chemokine receptor expression profile of hMSCs is wide-ranging and often contradictory,19, 20, 23 which may reflect varying culture conditions, methodology, detachment methods, antibodies, and passage number. Indeed, we found that with prolonged passaging to passage 6, the hMSC growth rate slowed considerably, coupled with a change to a flattened morphology. We observed variations in adhesion molecule expression with passage (Fig. 2A,B). CD49b expression significantly increased with passage number, which may indicate a more committed cell population, as CD49b interaction with collagen is required for osteoblast-specific gene expression.24 A decrease in CD44 expression was seen at passage 6, which has also been associated with differentiation or lineage specification.25 These changes, coupled with the change in growth rate and morphology, led us to standardize the use of hMSCs in our studies such that they are all passage 4. However, in support of our findings, CD29 and CD44 are consistently reported as highly expressed on hMSCs in terms of percentage and MFI.1, 26 Our findings of undetectable CD18 and CD61 are interesting because there is controversy regarding their reported levels on hMSCs,23 and their respective ligands ICAM-1 and CD31/PECAM-1 are highly expressed within the liver. CD44 was highly expressed on our hMSCs, with expression of both standard and variant isoforms of CD44 having been reported.26 We observed low levels of surface chemokine receptor expression, which is consistent with the findings of others.27

Although CD29, CD44, and chemokine receptors are important for recruitment of other cell types to the liver,15, 16, 28 to date no studies have addressed the mechanisms governing hMSC adhesion. Our study demonstrates preferential binding of hMSCs to injured liver, which is most likely due to up-regulation of adhesive proteins (e.g., VCAM-1, fibronectin) on HSECs and within extracellular matrix as we have described previously.15 This is supported by our observations that CD29 and VCAM-1 blockade mainly reduced binding of hMSCs to diseased liver. VCAM-1 is expressed on portal tracts in normal liver and on portal tracts and parenchyma in significant quantities in diseased liver.15 We found that blockade of CD44 on both hMSCs and HSECs15 reduced sinusoidal hMSC binding in normal and diseased tissue. This can be explained in part by the fact that the CD44 ligand hyaluronan is highly expressed in liver endothelium under both basal and inflammatory conditions.16, 29 Expression of hyaluronan is induced by proinflammatory stimuli in many types of microvascular endothelium; however, it is unclear whether inflammatory conditions affect hyaluronan expression within the liver.30 These data suggest the existence of constitutive (CD44/hyaluronan) and inducible (CD29/VCAM-1) mechanisms by which hepatic adhesion of hMSCs is regulated, raising the possibility of differential functions of hMSCs in the liver in physiological versus injury states.

In order to gain access to stromal compartments within tissue, it is necessary for blood-borne hMSCs to adhere to the endothelial lining of vessels. Using an assay designed to model such interactions under appropriate shear stress, we found that hMSCs were captured by endothelial-expressed VCAM-1 in a fast-rolling interaction that was inhibited by blocking either CD29 or VCAM-1. Previous studies have shown that hMSCs undergo P-selectin–dependent14 rolling on human umbilical vein endothelial cells (HUVECs); however, no role for selectins in leukocyte recruitment across hepatic sinusoids has been reported.31, 32 Furthermore, P-selectin is not expressed on HSECs due to the lack of Weibel-Palade bodies.33, 34 Our data therefore suggest parallels between leukocyte and hMSC capture by HSECs, with integrin-mediated interactions being vital. Our rolling interactions were faster than those reported for other cell types but were in agreement with previous hMSC adhesion studies using immobilized VCAM-1(250 μm/second) and HUVECs (400-500 μm/second).14, 35 We found hMSCs rolled on VCAM-1 at approximately 450μm/second and on HSECs at approximately 500 μm/second. Rolling velocities are affected by shear stress,36 density of receptor expression,37 and integrin activation state.38 Thus the rapid rolling that was observed may reflect lower densities of VCAM-1 on HSECs than other endothelium and donor-dependent variation in hMSC integrin activation status. Although our flow assay system incorporates flow channels much deeper than hepatic sinusoids, we attempted to model “physical trapping” promotion of firm adhesion by stopping flow for 5 minutes to allow hMSC-HSEC interactions. This modified assay confirms the central role for CD29, CD44, and VCAM-1 in the control of hMSC adhesion in our in vitro/ex vivo human systems as well as in our in vivo murine liver injury model.

Interestingly, combination block did not reduce adhesion any further than individual CD29 or CD44 blockade. Indeed, a physical association between CD44 and VLA-4 on hMSCs has been reported such that when hMSCs engage hyaluronan, CD44 and VLA-4 coassociation is markedly increased.35 Our combination block data support this observation and suggest that the two receptors may work together as a complex to regulate firm adhesion. However, our observation that only CD29/VCAM-1 regulates rolling suggests that the CD44/hyaluronan axis is not involved in this process. We propose, therefore, that in uninjured tissue, hMSCs bind constitutively expressed hyaluronan through passive adhesion. Increased VCAM-1 during liver injury allows rolling of hMSCs, thus driving active and passive adhesion to hyaluronan, fibronectin, and VCAM-1, thereby increasing adhesion.

It has been reported that physical trapping of large hMSCs in narrow sinusoids may precede engagement of firm adhesion receptors and transmigration into tissue.13 Our blocking experiments are key in this regard in that they demonstrate a reduction in adherent hMSCs in mouse liver, suggesting that active (in this case CD29 and CD44) processes operate in addition to any physical trapping that may occur. This passive entrapment has been ascribed to the large size and deformability of hMSCs. Moreover, our portal infusion route may enhance physical trapping, and because hMSCs are heterogeneous in size,39 it is possible that larger hMSCs undergo passive trapping in the sinusoids. However, we found that blockade reduced adhesion of both smaller and larger cells to the same extent, suggesting that even if physical trapping of hMSCs occurs, firm adhesion within the sinusoids requires active processes.

Our use of PTX was designed to assess whether G-protein–coupled receptors such as chemokine receptors could mediate hMSC adhesion to the liver. We found no reduction in hMSC binding with PTX treatment despite the cells' response to chemokine receptor ligands in chemotactic assays. Other cell types require engagement of chemokine receptors for integrin activation and subsequent migration.40 However, it appears that they are not critical for initial interactions between hMSCs and the liver. It is possible that although chemokines are not involved in adhesion, they are involved in transendothelial migration (TEM). Indeed, Thankamony and Sackstein35 demonstrated that G-protein–coupled receptors are involved in TEM of hMSCs using PTX. We did not observe TEM in the timeframe of our in vitro or in vivo assays, although this may reflect time taken for hMSCs to undergo TEM. Whereas leukocytes can undergo TEM after 5-20 minutes41 which can be seen in the timeframe of a flow adhesion assay, hMSCs have been reported to take 240 minutes before integrating into the endothelium.42 Moreover, hMSCs may use chemokine receptors to determine their position within the hepatic parenchyma near sites of inflammation and injury. Whereas the CCl4 model used to induce injury in our mice would have resulted in up-regulation of a similar cytokine and chemokine profile to that seen in aspects of human disease,43 other cytokine combinations may exert different responses in other models of injury that require chemokine receptors for hMSC interactions.

Understanding the factors controlling the ingress of hMSCs to injured liver will allow for their targeting, so as to improve/reduce the number of hMSCs arriving there. The impact of changing hMSC number in injured liver is yet to be elucidated, partly due to limitations of xeno-transplantation. Immunogenicity in xeno-transplantation models is likely to be an issue given that even allogeneic MSCs are rejected.44 Similarly, use of immunocompromised mice is significantly limited as the immunomodulatory properties of hMSCs3-5, 7 are best tested in models of immune-mediated liver injury. Studies assessing the impact of changing MSC numbers in injured organs will therefore require further investigation using syngeneic and allogeneic murine MSCs in appropriate mouse models of liver injury. That we observe increased hMSC interaction/egress with injured liver would suggest that this axis is functionally relevant as supported by studies in other organ settings.12

Clinical trials are ongoing in which hMSCs are being used in graft-versus-host disease3, 4 and there is potential for use of hMSCs as vectors for antitumor therapy.19 Understanding how hMSCs engraft in the injured liver will be critical in the design of new therapeutic approaches, whether it be to augment their ingress or prevent it. Sackstein et al.27 demonstrated that glycosylation of CD44 can improve engraftment of hMSCs to bone marrow, while Kumaran et al.45 demonstrated that infusion of soluble FLP into liver improves hepatocyte engraftment. These approaches illustrate the scope to improve the efficiency of cellular therapy, and in this study we identified CD29 and CD44 as promising candidates that may be engineered prior to introduction into patients to control delivery and retention of hMSCs within the liver.