No Evidence for Significant Transdifferentiation of Bone Marrow Into Pancreatic β-Cells In Vivo

Animals were maintained under conditions approved by the institutional committee on research animal care. Female C57BL/6 (B6, H-2^b; age 10 weeks) and male actin green fluorescent protein (GFP)-transgenic B6 mice (C57BL/6-TgN[ACEbEGFP]10sb; age 6 weeks) were purchased from Jackson Laboratory (Bar Harbor, ME) and used as the recipients and donors, respectively. In two independent experiments, recipient mice were lethally irradiated (10 and 11 Gy, respectively; ¹³⁷Cs source; 0.8 Gy/min) and reconstituted within 4–8 h by a single intravenous injection of 10 × 10⁶ bone marrow cells from GFP⁺ donors. Tissue analysis was performed 3 months after (experiment 1) or 4–6 weeks (experiment 2) after bone marrow transplantation (BMT). The percentage of GFP⁺ cells in the recipient blood was determined by flow cytometric analysis 4 weeks and 3 months (experiment 2 only at 4 weeks) after BMT and levels of donor chimerism were calculated by the following formula: donor chimerism (%) = (% GFP⁺ cells of the BMT recipient/% GFP⁺ cells of normal GFP-transgenic mice) × 100. At the end of the study, the mice were killed by CO₂ asphyxiation.

Mice were anesthetized with a combination of isoflurane and pentobarbital. After an abdominal midline incision, the pancreas was exposed and 60–70% of pancreatic tissue was gently removed with a cotton swab. Only tissue in the immediate vicinity of the duodenum was left behind. The abdominal wall was closed with 5.0 silk, and wound clips were applied to the skin. Animals were killed 7–14 days after the procedure.

Streptozotocin (STZ)/insulin was administered as previously described by Guz et al. (23). Mice were injected with 200 mg/kg STZ i.p. after an overnight fast. Blood glucose was measured 24 h later to confirm hyperglycemia >200 mg/dl. The animals then received a subcutaneous implant of two LinBit sustained insulin release pellets (LinShin, Scarborough, CA), given according to the manufacturer’s instructions. Animals were killed 12 days after STZ injection.

Pancreatic tissue was fixed overnight in 10% buffered formalin, embedded in paraffin, and cut into 4-μm sections. Sections were dewaxed in xylene, hydrated, and boiled for 10 min in a microwave oven in 10 mmol/l sodium citrate (pH 6) for antigen retrieval. Sections were blocked with PBS/normal donkey serum, incubated with first antibodies at 4°C overnight, washed, incubated with Cy2- or Cy3-labeled secondary antibodies (Jackson ImmunoResearch, West Grove, PA), washed again, and mounted with fluorescence mounting medium. Antibodies used were rabbit and mouse anti-GFP (1:300; both from Chemicon, Temecula, CA), goat anti-amylase (1:100; Santa Cruz, Santa Cruz, CA), guinea pig anti-insulin (1:5,000), guinea pig anti-glucagon (1:2,000), rabbit anti-somatostatin (1:2,000), and guinea pig anti-pancreatic polypeptide (1:2,000; all from Linco, St. Charles, MO).

For immunostaining of platelet endothelial cell adhesion molecules (PECAM; 1:100; goat; Santa Cruz) and CD45 (1:50; rat; BD Pharmigen), pancreatic tissue was fixed for 3–5 h in 10% buffered formalin, cryoprotected in 30% sucrose/PBS overnight at 4°C, and frozen in optimal cutting temperature compound. Next, 5-μm cryosections were stained as described above, but without antigen retrieval and using PBS/0.1% Triton instead of pure PBS.

Images were acquired with a Spot digital camera (Diagnostic Instruments, Sterling Heights, MI). Digital overlay for multicolor images was done with Adobe Photoshop 6.0.

After partial pancreatectomy, some animals were injected intraperitoneally with bromo-deoxyuridine (BrdU) (0.1-ml cell proliferation labeling reagent; Amersham Biosciences, Little Chalfont, U.K.) 12, 36, and 60 h and 4, 6, 9, and 12 days after the intervention. To visualize BrdU staining, paraffin-embedded tissue sections were hydrated and incubated in 2 mol/l HCl and in 0.1 mol/l HCl plus 0.05 mg/ml pepsin (Sigma, St. Louis, MO), each kept at 37°C for 20 min. Immunostaining was done with a mouse monoclonal anti-BrdU antibody (1:300; Sigma) overnight at 4°C and the mouse-on-mouse detection kit (Vector, Burlingame, CA) using NovaRed (Vector) as a peroxidase substrate.

To approximate the number of β-cells that we microscopically examined in this study, we immunostained several pancreas sections for insulin, measured the surface area of the insulin-stained β-cell core of randomly selected islets, and counted the number of β-cell nuclei (nuclear holes clearly surrounded by insulin-positive cytoplasm) in these islets. We next calculated the average number of β-cell nuclei per 1,000 μm² of islet area (4.76). We then measured the surface area of all the insulin-stained islets on one representative section each from all study animals and used these measurements to calculate the average β-cell number per tissue section for each animal. For STZ-administered animals, all four major endocrine islet cells were included in this calculation instead of β-cells. The numbers in Table 2 represent the average number of β-cells per tissue section multiplied by the number of sections that were examined. The number of analyzed amylase-positive exocrine cells was estimated in a similar fashion (1.49 cells/1,000 μm²; average number of cells per tissue section: 42,000).

After being immunostained and captured as digital images, selected tissue sections were further processed for Y chromosome fluorescent in situ hybridization (FISH). The coverslip was removed and slides washed once in 2× sodium chloride−sodium citrate (SSC) and then incubated in 1 mol/l sodium thiocyanate for 30 min at 80°C. Slides were then washed twice in 2× SSC and digested in proteinase K (10 μg/ml in 2× SSC; Sigma, St. Louis, MO) for 90 min at 37°C, washed again in 2× SSC, air dried, and treated with methanol/acetic acid 3:1 for 5 min at room temperature. After being air dried, slides were denatured at 72°C for 10 min in 70% deionized formamide/2× SSC. Denaturation was stopped for 2 min in cold (−20°C) 70% ethanol. Slides were dehydrated in 90 and 100% ethanol for 2 min each and air dried. Next, 12 μl of the biotinylated mouse Y chromosome probe (Cambio, Cambridge, U.K.) were applied under a sealed coverslip overnight at 37°C. Probe denaturation, posthybridization washes, and signal detection with the Biotin Texas Red Amplification Painting Kit (Cambio) were done following the manufacturer’s instructions. Sections were mounted in fluorescence mounting media containing 4′,6-diamidino-2-phenylindole (DAPI) as a nuclear counterstain.

To address the question of whether bone marrow cells can give rise to pancreatic endocrine cells, in particular β-cells, we lethally irradiated female wild-type C57BL/6 mice and then transplanted them with unfractioned bone marrow from syngeneic, male, actin GFP-transgenic mice. In these animals, GFP is expressed under the control of the chicken β-actin promoter plus the cytomegalovirus enhancer, and all tissues with the exception of erythrocytes and hair are GFP⁺ (24). We performed two independent experiments with tissue analysis 3 months (experiment 1) and 4–6 weeks (experiment 2) after BMT (Fig. 1A). At 4 weeks after BMT, all animals in both experiments showed high donor chimerism of peripheral blood nucleated cells (mean 89.5%; range 84.8–92.9%), which was further increased at 3 months after BMT in experiment 1 (mean 96.6%; range 91.2–99.9%) (Fig. 1B; Table 1).

At 2 weeks before the end of each experiment, we separated the mice into three groups. One group underwent partial pancreatectomy, one group received STZ and subcutaneously implanted insulin pellets, and the third group (control) went without any further intervention (Fig. 1A).

We next established that pancreatic tissue, including β-cells, in the actin-GFP donor strain of mice was homogenously GFP⁺ and that we could clearly distinguish positive from negative tissue by immunofluorescent staining (Fig. 2A and B). In addition, we confirmed that all cells in freshly isolated islets from the donor strain were robustly GFP⁺ (data not shown). We also tested FISH for the mouse Y chromosome on male and female control tissue (Fig. 2C). Using 4-μm paraffin sections, we detected a Y chromosome in ∼60% of male nuclei. This percentage is explained by the fact that a single tissue section often contains only part of an individual nucleus, which may or may not contain the Y chromosome.

To investigate whether bone marrow contributes to β-cell turnover under steady-state conditions, we examined pancreatic tissue from mice that were killed 4–6 weeks or 3 months after BMT without any further intervention. We analyzed multiple tissue sections from all animals by fluorescent doublestaining with antibodies against GFP and insulin. We inspected >30,000 individual β-cells in each of the two experiments. In the animals killed after 4–6 weeks, we found no cells that were co-stained for insulin and GFP. After 3 months, we detected two such cells, one each in two different animals (Fig. 3A; Table 2). Co-localization of the two proteins in the same cell was confirmed by confocal microscopy (Fig. 3E); however, Y chromosome staining of both cells was negative (data not shown). We also detected three GFP⁺ pancreatic exocrine cells after 3 months (Fig. 3B and F), but no co-localization of GFP and the other major islet hormones (i.e., glucagon, somatostatin, and pancreatic polypeptide; data not shown).

In contrast to the low donor contribution to endocrine and exocrine cells, the pancreases of all animals contained large numbers of donor-derived hematopoietic (CD45⁺) and interstitial cells (Fig. 3C and D, G–M). Most interstitial GFP⁺ cells did not co-stain with PECAM, a marker of vascular endothelial cells (Fig. 3L). We only occasionally observed donor-derived vascular endothelial cells (Fig. 3M).

We also examined several tissue sections by Y chromosome FISH as a transgene-independent technique to detect donor tissue; we found no GFP-negative/Y chromosome−positive β-cells. By digital overlay of FISH and immunostaining images, we were, however, able to demonstrate a Y chromosome in the majority of GFP⁺ cells (Fig. 3I and J).

We next determined if pancreatic injury would increase the number of bone marrow−derived cells that adopted an endocrine phenotype. We therefore performed a 60–70% pancreatectomy in several mice 7–14 days before killing the animals. BrdU labeling demonstrated the proliferation of ductal and islet cells after pancreatectomy (Fig. 4A and B). We also observed typical regenerative foci (21), with multiple small ducts and islets surrounded by interstitial tissue. We found that in these foci a high fraction of interstitial cells, but no islet or duct cells, were of bone marrow origin (Fig. 4C and D). Analysis of pancreatic tissue sections spaced throughout the depth of the pancreatic remnants revealed no GFP⁺ or Y chromosome−positive β-cells or other endocrine cells in any of the pancreatectomized animals (Table 2). We found one GFP⁺ exocrine cell in one mouse 3 months after BMT.

As a second model of islet cell injury, we admin-istered 200 mg/kg STZ i.p. to bone marrow−transplanted mice, then subcutaneously implanted insulin-releasing pellets. Insulin treatment has previously been shown to support the partial restoration of β-cell mass in STZ-diabetic CD1 mice (22,23). We found in pilot studies with female C57BL/6 mice, similar to those used in this BMT project, that the proliferation of endocrine non–β-cells was a prominent feature of islet remodeling after STZ in this mouse strain. The same was true for the bone marrow−transplanted mice in this study. When the animals were killed 12 days after STZ administration, the absolute number of α- and δ-cells and PP cells per islet cross section was significantly increased over that of control mouse islets (Fig. 4E). We also found that a subgroup of cells co-expressed pancreatic polypeptide and another islet hormone such as insulin or glucagon, a type of cell not seen in normal adult mouse islets (Fig. 4F). Because many islet cells after STZ administration were non–β-cells, we used doublestaining with anti-GFP and an antibody cocktail of the four major islet hormones to look for donor-derived endocrine cells. However, no donor-derived endocrine cells were found in any of these animals, whereas the number of interstitial cells of donor origin was comparable with that of control animals (Fig. 4G, Table 2).

In this study, we found no significant in vivo differentiation of bone marrow into pancreatic islet β-cells in adult mice, neither under steady-state conditions nor after tissue injury through partial pancreatectomy or STZ administration. We analyzed pancreatic tissue 4–6 weeks and 3 months after BMT. No donor marker−positive β-cells were found at 4–6 weeks, whereas we detected two positive cells at 3 months (0.004% of all β-cells examined). Both of the cells were in control mice that had not undergone additional pancreatic injury.

In contrast, all animals exhibited stable and almost complete hematopoietic donor chimerism. Moreover, tissue analysis of the pancreas revealed a high contribution of transplanted bone marrow cells to cells in the pancreatic interstitium and to blood and lymphatic cells. These findings indicated that the host bone marrow was efficiently repopulated with donor cells and that the lack of donor-derived β-cells was not caused by incomplete chimerism.

The analysis of pancreatic tissue sections as well as isolated islets from the donor strain demonstrated robust GFP expression in β-cells. Therefore, it seems unlikely that GFP would be specifically downregulated in donor-derived β-cells after BMT. We also applied Y chromosome FISH as a transgene-independent method to identify donor-derived cells. Using this method, we detected no additional bone marrow−derived β-cells. We are therefore confident that GFP immunostaining did not significantly underestimate the number of marrow-derived β-cells in our study, even taking into account that the sensitivity of Y chromosome FISH on the 4-μm tissue sections used in this study is only about 60%. We were unable to detect a Y chromosome in the two GFP⁺ β-cells found and thus could not confirm the donor origin of these cells with complete certainty. However, based on a 60% sensitivity, the chance of missing two subsequent Y chromosomes is 16%. Lacking an alternative explanation for GFP expression in these cells, we therefore still believe that donor cells contributed to the genome of the GFP⁺ β-cells by either transdifferentiation or fusion (see below).

We also studied two established models of pancreatic injury repair. Tissue regeneration after partial pancreatectomy has been investigated by several groups over more than two decades and is believed to involve duct-derived progenitor cells (21,25–27). Regeneration after partial pancreatectomy is rapid, with the main proliferative response occurring during the first week (21). Islet regeneration in mice after STZ administration has been demonstrated more recently by two groups (22,23,28); further, insulin supplementation has been shown to have a positive effect on regeneration (23). Using these models, we did not observe any bone marrow contribution to pancreatic epithelial tissue. We have provided evidence for duct and islet cell proliferation after partial pancreatectomy and for islet remodeling after STZ administration, but because of the limited number of bone marrow−transplanted animals, we did not quantify the extent of tissue regeneration in this study. Therefore, it remains possible that other experimental models can induce more efficient pancreas regeneration with a bone marrow contribution that was not seen in our work.

Our results differ from those of a recent study by Ianus et al. (20), who found that 1.7–3% of islet β-cells were of bone marrow origin 4–6 weeks after bone marrow transplantation in mice. The reasons for these different outcomes are unclear at this point.

Both our study and the one reported by Ianus et al. (20) were done in the same strain of mouse. The dosages of radiation used for host preconditioning (10/11 Gy in our study vs. 10.5–11 Gy in the Ianus et al. study) were similar, but the number of transplanted bone marrow cells was 10 times higher in our study (10⁷ vs. 10⁶ cells). Peripheral blood donor chimerism was also stronger in our study. Tissue analysis was done 4–6 weeks after BMT in both studies and also 3 months after BMT in ours.

Ianus et al. (20) used an insulin promoter−based Cre-Lox/GFP system to quantify the number of donor-derived β-cells in contrast to the constitutively GFP-expressing bone marrow used in our investigations. Whether this difference could explain the contrasting results remains to be determined. Another possible explanation could be unrecognized differences in the preconditioning of the mice for BMT or the preparation of the donor cells. Hypothetically, a yet-unknown type of stem cell that can give rise to pancreatic β-cells but that is only inconsistently replaced by donor cells with the current BMT regimens, could exist within the bone marrow. Even in the presence of full hematopoietic donor chimerism, such a cell type might then still be of host origin and its progeny would not be recognized as being derived from bone marrow.

In another recent report, Hess et al. (29) found that the transplantation of bone marrow cells after diabetes induction with STZ supports the recovery of pancreatic β-cell mass and partial reversal of hyperglycemia. This study noted that after STZ administration, up to 2.5% of insulin-positive cells were also donor marker positive. Yet these cells did not express the transcription factor PDX1, usually seen as an essential feature of functional β-cells. Donor-derived, insulin-positive cells also did not account for the functional recovery seen in the study. Instead, BMT augmented the regeneration of host β-cells by a yet-unknown mechanism. This could involve bone marrow−derived endothelial progenitor cells, as up to 9% of donor cells in the pancreases of STZ-administered animals in this study stained positive for the vascular marker PECAM. We did not observe a similarly high frequency of donor-derived vascular cells, but only occasional cells that co-stained for PECAM. However, the two experimental approaches are not entirely comparable: we established stable bone marrow chimerism first and then inflicted pancreatic injury, whereas Hess et al. gave bone marrow after STZ as a therapeutic intervention. This approach could bring a number of vascular progenitor cells directly to the injured pancreas, more than what the bone marrow provides regularly.

Our study did not address the question of whether the small number of donor marker−positive β-cells (two) was the result of true transdifferentiation or cell fusion. Cell fusion was recently found to be the primary mechanism by which bone marrow cells rescue mice with fumarylacetoacetate hydrolase deficiency, a hereditary defect in liver metabolism (18,19). Our understanding of in vivo cell fusion is still minimal. However, one prerequisite seems to be that the resulting tetraploid cell can be functional and is not forced to undergo apoptosis. The naturally high frequency of tetraploid hepatocytes might therefore support the successful fusion of some bone marrow−derived cells with liver cells. Most pancreatic β-cells in vivo are diploid, but β-cells with more than two sets of chromosomes are present in human and rodent islets (30,31). Thus it seems possible that a low rate of fusion between bone marrow−derived cells and β-cells occurs in vivo. This could be an alternative explanation for the very low numbers of pancreatic β-cells that expressed bone marrow markers. The fact that we were able to find two GFP⁺ β-cells 3 months, but not 4–6 weeks, after BMT could indicate that the low level contribution of bone marrow to pancreatic tissue is a slow process. Yet, given the overall small number of positive cells, a sampling error can also not be excluded.

Differing results on the differentiation potential of bone marrow stem cells are not unique to the field of diabetes; although two groups have reported the in vivo differentiation of central nervous system neurons from transplanted bone marrow (11,32), another study using a similar experimental approach had a negative outcome (17). Krause et al. (33) described how transplantation with a single bone marrow cell could lead to engraftment in multiple different tissues in addition to full hematopoietic reconstitution, yet others were unable to reproduce these findings (16). Finally, as mentioned above, the differentiation of bone marrow cells into hepatocytes (13) has recently been identified as cell fusion, not independent differentiation (18,19).

The results of present study support the notion that the differentiation capacity of adult bone marrow stem cells is limited and that pancreatic β-cells are not substantially derived from bone marrow in vivo. However, at this point our results are only applicable to the mouse models used in this study. Whether β-cells are derived from bone marrow in other species or experimental settings needs to be determined. Our studies also did not exclude the possibility that β-cells could be derived from bone marrow stem cells in vitro. Results with marrow-derived MAPCs (15) and brain-derived oligodendrocyte precursors (34) suggest that the differentiation capacity of cells in vitro can be broader than in vivo. On the other hand, a recent report that insulin uptake can mimic β-cell differentiation in vitro cautions that such experiments need to be done carefully (35).

The conflicting results in other areas of adult stem cell biology remind us that this field of investigation is still at an early stage. Further studies are necessary to establish definitive answers on the versatility of different adult stem cell populations and their utility for the treatment of disease. In particular, new techniques to trace the progeny of adult stem cells in vivo have to be developed along with clonal in vitro assays that test the differentiation capacity of various cell types.

During revisions of this manuscript, a report was published by Choi et al. (36) also describing the lack of significant bone marrow–to–pancreatic β-cell transdifferentiation in bone marrow−transplanted mice.

Support for this study was provided by U.S. Public Health Service Grant DK-60125 (J.F.H.), Juvenile Diabetes Research Foundation International Grant 1-2001-770 (Y.-G.Y.), and the Deutsche Forschungsgemeinschaft (A.L.). J.F.H. is an investigator with the Howard Hughes Medical Institute.