An uncomfortable silence … while we all search for a better reporter gene in adult stem cell biology


  • Ryan A. McTaggart M.D.,

    1. UCSF Liver Center (P30 DK26743) and Department of Surgery, Division of Transplantation, University of California, San Francisco, San Francisco, CA
    Search for more papers by this author
  • Sandy Feng M.D., Ph.D.

    1. UCSF Liver Center (P30 DK26743) and Department of Surgery, Division of Transplantation, University of California, San Francisco, San Francisco, CA
    Search for more papers by this author

Terai S, Sakaida I, Yamamoto N, Omori K, Watanabe T, Ohata S, et al. An in vivo model for monitoring trans-differentiation of bone marrow cells into functional hepatocytes. J Biochem (Tokyo) 2003;134:551–558. (Reprinted with permission.)


The plasticity of bone marrow cells (BMCs) remains controversial. The present study found that persistent injury induces efficient trans-differentiation of BMCs into functional hepatocytes. Mice with liver cirrhosis induced by carbon tetrachloride were injected with 1 × 10(5) non-treated green fluorescent protein (GFP)-positive BMCs via the tail vein. In these mice, transplanted GFP-positive BMCs efficiently migrated into the peri-portal area of liver lobules after one day, repopulating 25% of the recipient liver by 4 weeks. In contrast, no GFP-positive BMCs were detected following transplantation into control mice with undamaged livers. BMCs trans-differentiated into functional mature hepatocytes via immature hepatoblasts. Serum albumin levels were significantly elevated to compensate for chronic liver failure in BMC transplantation. These results reveal that recipient conditions and microenvironments represent key factors for successful cell therapy using BMCs.


Stem cells —self-renewing and multipotent—are, in principle, ideal for tissue renewal and correction of genetic defects. For decades, investigators have believed that certain tissues harbor stem cells capable of differentiating into cell types characteristic of that tissue. The ability to isolate and amplify such cells may represent a strategy to treat disease. In some tissues such as the liver, however, the notion of a tissue-specific stem cell has been based more on deductive reasoning than on actual observation. Cells with these characteristics have never been clearly identified nor isolated. Hence the intense interest when, in 1999 and 2000, two groups independently reported that liver stem cells may reside in the adult bone marrow (BM).1, 2 These and other subsequent experiments involved transplantation of “marked” BM cells. Examination of the liver revealed hepatocytes bearing these markers, suggesting that cells originating from the BM can engraft within the liver and differentiate into hepatocytes. While models with mild or no liver injury resulted in a few BM-derived hepatocytes,1–3 a model with lethal liver injury resulted in massive liver replacement by BM-derived hepatocytes.4 Observational studies in humans appeared to confirm these findings in murine systems.5 A new era in which tissue-specific stem cells may no longer be needed appeared to be at hand.

The initial optimism was, however, followed by retrenchment. Several subsequent reports, taken together, suggested that liver engraftment by BM cells may be so infrequent as to have little therapeutic utility (Table 1). These studies consistently observed either a negligible contribution or no contribution of donor BM cells to liver regeneration in models with or without liver injury.1–3, 6–8In toto, they suggested that the transformation of a BM cell into a liver cell requires hematopoietic reconstitution, occurs rarely and slowly, and cannot be readily enhanced under physiological conditions.1–3, 6–9 The few models yielding substantial liver replacement likely resulted from strong in vivo selection of rare transformation events of donor BM cells that harbored a survival advantage over native recipient cells.4, 7, 9–11

Table 1. Murine Liver Engraftment by Hematopoietic Stem Cells
PublicationYearDonor Marker (Reporter)Recipient TreatmentDonor Hepatocytes (%)Time After Injury
BMTxLiver Injury
  • Abbreviations: BMTx, bone marrow transplant; CD26, dipeptidyl peptidase IV; 2-AAF, 2-acetylaminofluorene; LacZ, ROSA26-β-galactosidase; FAH, fumarylacetoacetate hydrolase; hAlbumin, human albumin; hCK19, human cytokeratin 19; NR, not reported; HBsAg, hepatitis B surface antigen; CDE, choline deficient, ethionine-supplemented diet.

  • *

    Time after BMTx.

  • BM injected without total-body irradiation.

Petersen et al.11999Y-chromosome or CD26Yes2-AAF + CCl40.169 and 13 days
Thiese et al.22000Y-chromosomeYesNone2.224 weeks*
Lagasse et al.42000LacZYesFAH deficiency30–5028 weeks
Wagers et al.62002GFPYesNone<0.000136 weeks*
Mallet et al.72002Bcl-2YesFas-agonist antibody0.05–0.88 weeks
Wang et al.92002LacZYesFAH deficiencyNone5 weeks
     30–5022 weeks
Wang et al.172003hAlbumin or hCK19YesCCl4NR4 weeks
Terai et al.122003GFPNoCCl4264 weeks
Kanazawa et al.82003GFP or LacZYesCCl4None4 and 8 weeks
  Y-chromosomeYesCCl4<0.00014 and 8 weeks
  GFP or LacZNoCCl4None4 and 8 weeks
  GFPYesUrokinase expressionNone15 weeks
  GFP or LacZYesHBsAg expressionNone13–32 weeks
  GFPYesHBsAg expression + CDENone47 weeks


Figure 1.

5 μm liver section from a mouse after 5 weeks of twice-weekly intraperitoneal injections of carbon tetrachloride (0.5ml/kg). Tissue section was stained with DAPI (blue), coverslipped, and then viewed with an epifluorescent microscope. “Green” cells surrounding this central vein correspond to dead/dying hepatocytes that, when viewed with light microscopy, stain intensely with eosin (acidophil bodies). This mouse was not transplanted with GFP BM; the mouse was not irradiated and did not receive BM from a GFP transgenic mouse. Abbreviation: CV, central vein.

In November 2003, Terai et al. presented data at odds with the view that liver engraftment by BM-derived cells is slow and impractical.12 They suggested that adult BM cells, infused into mice with cirrhosis induced by chemical liver injury, can transdifferentiate into hepatocytes within 1 day and by 4 weeks repopulate as much as 26% of the liver. What might explain this apparent contradiction in the frequency and/or efficiency of BM contribution to liver regeneration?

Terai and colleagues used adult male transgenic mice (act-EGFP) engineered to constitutively express green fluorescent protein (GFP) under the control of the chicken β-actin promoter as BM donors for female C57BL6 mice that had received 4 weeks of twice-weekly intraperitoneal injections of carbon tetrachloride (CCl4) at 0.5 mL/kg. Bulk GFP-BM cells were injected intravenously without prior radiation, and CCl4 treatment was continued. The liver was assessed for GFP-expressing cells by immunohistochemistry and fluorescence microscopy beginning 1 day after BM infusion and then at weekly intervals. At 4 weeks, no GFP-positive hepatocytes were detected in mice without either liver injury or GFP-BM infusion. However, in mice with liver injury, GFP-positive hepatocytes were observed 1 day after GFP-BM infusion and proliferated over the next 4 weeks to account for 26% ± 1% of all hepatocytes. Notably, cells expressing Liv2 and HNF4, two markers of hepatocyte differentiation, also increased during the four weeks after BM injection. Epifluorescence microscopy demonstrated cells coexpressing GFP and either albumin, Liv2, or HNF4. In contrast, oval cells, observed by 1 week after BM infusion, failed to increase in number. The function of BM-derived hepatocytes was inferred by improvement in serum albumin. Terai and colleagues concluded that BM transplantation without preconditioning can contribute to liver regeneration in the setting of severe chronic liver injury/cirrhosis. Furthermore, they suggest that this robust and efficient regeneration, efficacious in ameliorating disease, occurs by transdifferentiation.

The findings of Terai and colleagues are certainly intriguing. Can donor BM cells without survival advantage contribute efficiently and substantially to liver regeneration if given to animals undergoing chronic chemical liver injury (CCl4) but without preconditioning?

Terai and colleagues report that BM-derived hepatocytes appear first in periportal regions and then stream out into the hepatic lobule. However, the immunohistochemistry and fluorescence images at the magnification displayed in the publication do not conclusively demonstrate either the expected centrizonal injury that CCl4 is known to induce or the periportal location of GFP-positive cells. The initial periportal location (Zone 1) is important for two reasons. First, previous studies of transplanted hepatocytes and BM cells consistently point to the periportal region as the preferential port of cellular entry.13, 14 Second, the injury induced by chronic CCl4 administration is known to target centrizonal areas (Zone 3) with relative sparing of portal areas.15, 16 This differential effect may represent a “selection mechanism” for the preferential survival of periportal cells, either native hepatocytes and/or BM-derived cells.

The authors also describe that BM-derived cells were observed “spreading into the liver lobules 1 to 4 weeks after transplantation … and forming liver cell cords” during ongoing CCl4 injury. This is also novel, since previous demonstrations of liver repopulation by BM cells have noted either isolated BM-derived cells1, 2, 6–9, 17 or clusters/nodules.4, 9–11 Single cells demonstrate “proof of principle,” the transformation of a BM cell into a hepatocyte, but they question the replicative potential of the resulting BM-derived hepatocyte. Clusters or nodules, even if rare, strongly suggest the replication competency of the BM-derived cells. Terai et al. note differences in the architectural arrangement of their putative BM-derived cells compared with others but do not discuss it further. They do not provide information regarding the relationship among their observed GFP-positive cells. Do they represent individual and thus highly frequent “transformation” events? Or are there multiple clonal families? Or both? Does the architectural pattern change over time with further injury? And what implications do the answers to the above questions have regarding the replicative competency of the GFP BM-derived cells? Perhaps there are no clusters or nodules because GFP-positive cells are replication deficient or replication compromised. GFP interferes with the normal biology of cellular differentiation and proliferation.18In vitro studies have reported free-radical phototoxicity after GFP excitation.19 Furthermore, Huang et al. have reported that transgenic mice with cardiac myocyte GFP expression develop fatal cardiomyopathy.20 These studies suggest that GFP expression may be neither biologically inert nor benign, two characteristics essential for a useful reporter gene.

Unambiguous identification, another essential characteristic of the ideal reporter gene, is poorly served by GFP, because hepatocytes autofluoresce. To make matters worse, hepatocyte injury and/or senescence may result in high-level expression of lipofuscin, flavins, and age-related proteins, which further increase autofluorescence.21–23 Like Terai, we used CCl4 to induce chronic liver injury in C57BL6 mice. Hematoxylin and eosin staining demonstrated the expected liver injury in centrizonal areas with hepatocyte necrosis, evidenced by abundant acidophil bodies and an accompanying inflammatory infiltrate. Fluorescence microscopy of frozen sections showed intensely green cells clustered around central veins that were also positive for hepatocyte-specific gene products such as HepPar1, albumin, and dipeptidyl peptidase IV. Comparison of these images suggest that the acidophil bodies corresponded to the fluorescent cells. We now exercise extreme caution in assigning cellular origin based simply upon fluorescence, as has been suggested by others.24

Aside from GFP, β-galactosidase and/or Y-chromosome detection have been popular strategies to demonstrate BM-driven tissue regeneration. Experiments in liver, muscle, and nerve regeneration have successfully used ROSA26 mice (constitutive β-galactosidase expression) to demonstrate tissue engraftment. Weak β-galactosidase expression at the single-cell level have led some investigators to reject ROSA26 mice as BM donors.25 Loss of β-galactosidase expression has also been reported: Transplantation of BM from male ROSA26 donors into lethally irradiated female mice resulted in 90% splenic engraftment by Y-chromosome analysis but <50% engraftment by X-gal staining.24 Detection of the Y-chromosome, perhaps the current gold standard, is expensive and labor-intensive, with sensitivity limitations imposed by the obligatory thickness of tissue sections. The lack of a rapid, robust, and reliable detection strategy for a marker gene or gene product poses a significant obstacle to progress in this research arena.

Most investigators believe that BM-derived hepatocytes occur but are rare. Thus, the findings reported by Terai et al. are notable. We eagerly await confirmation of their results and answers to the important questions that their results raise. Adult stem cell transplantation, embryonic stem cell biotechnology, and somatic cell nuclear transfer all hold promise for regenerative medicine. Currently, the last two require cumbersome and esoteric technology and are burdened by the spectre of undefined oncogenic, immunologic, and zoonotic risks. Since adult BM cells can, in principle, repopulate the liver and are readily available and routinely used in current clinical practice, we believe that understanding and improving the process remain elusive but worthy challenges.