编者按：骨髓干细胞，造血干细胞的转分化已争论了很长时间，先前说是有多向分化潜能，后来主要认为有三向分化潜能，即到骨，软骨和脂肪，后来又观察到向神经分化是可能的，并成功分化为神经系统的purkinje，再后来小鼠的bone marrow mesenchymal stem cell成功分化为cardiomyocytes，但未有其它group重复过。记得数年前有一篇Transdifferentiation is so difficult文章，便提示到这个问题，但未引起足够重视。而去年PNAS一篇研究内皮祖细胞转分化时发现，所谓转分化可能是由于cell fusion所致，尤其是核型鉴定尤其如此，便引发了大震荡，以前的成功分化的例子都不可靠？这篇文章以令人信服的证据表明，骨髓间质干细胞的转分化看来并不是很容易的！看来我们对事务的认识过程也是不段深化的。相似的，其它成年干细胞也是如此。当然，我们并不能因此对adult stem cells感到失望，上个月Cell报道cardiac stem cell成功地发现！而且能分化为多种细胞！总之，这种争论还将继续下去。（Bioon于2003-10-31)
Nature 425, 968 - 973 (30 October 2003); doi:10.1038/nature02069
Nature AOP, published online 12 October 2003
Fusion of bone-marrow-derived cells with Purkinje neurons, cardiomyocytes and hepatocytes
MANUEL ALVAREZ-DOLADO1, RICARDO PARDAL2, JOSE M. GARCIA-VERDUGO3, JOHN R. FIKE1, HYUN O. LEE2, KLAUS PFEFFER4, CARLOS LOIS5, SEAN J. MORRISON2 & ARTURO ALVAREZ-BUYLLA1
1 Department of Neurological Surgery, University of California at San Francisco, San Francisco, California 94143-0520, USA
2 Howard Hughes Medical Institute, Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan 48109-0934, USA
3 Instituto Cavanilles, University of Valencia, Valencia 46100, Spain
4 Institute of Medical Microbiology, University of Dusseldorf, D-40225 Dusseldorf, Germany
5 Picower Center for Learning and Memory, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139-4307, USA
Correspondence and requests for materials should be addressed to A.A-B. (abuylla@itsa.ucsf.edu).
Recent studies have suggested that bone marrow cells possess a broad differentiation potential, being able to form new liver cells, cardiomyocytes and neurons1, 2. Several groups have attributed this apparent plasticity to 'transdifferentiation'3-5. Others, however, have suggested that cell fusion could explain these results6-9. Using a simple method based on Cre/lox recombination to detect cell fusion events, we demonstrate that bone-marrow-derived cells (BMDCs) fuse spontaneously with neural progenitors in vitro. Furthermore, bone marrow transplantation demonstrates that BMDCs fuse in vivo with hepatocytes in liver, Purkinje neurons in the brain and cardiac muscle in the heart, resulting in the formation of multinucleated cells. No evidence of transdifferentiation without fusion was observed in these tissues. These observations provide the first in vivo evidence for cell fusion of BMDCs with neurons and cardiomyocytes, raising the possibility that cell fusion may contribute to the development or maintenance of these key cell types.
In order to detect cell fusion we used a method based on Cre/lox recombination, a technique extensively used to conditionally turn on or off gene expression in specific cell types or tissues, or at particular stages in development10. For this study we first used mice expressing Cre recombinase ubiquitously under the control of a hybrid cytomegalovirus (CMV) enhancer -actin promoter11 (Fig. 1a), and the conditional Cre reporter mouse line R26R12 (Fig. 1b). In this line, the LacZ reporter gene is exclusively expressed after the excision of a loxP-flanked (floxed) stop cassette by Cre-mediated recombination (Fig. 1b). When Cre-expressing (Cre+) cells fuse with R26R cells, Cre recombinase excises the floxed stop cassette of the reporter gene in the R26R nuclei, resulting in expression of LacZ in the fused cells. Consequently, fused cells can be detected easily by 5-bromo-4-chloro-3-indolyl--D-galactoside (X-gal) staining (Fig. 1c). This method previously failed to detect evidence of cell fusion in the pancreas13. For this reason, we first verified this cell fusion detection method in vitro.
Figure 1 Method to detect cell fusion events. Full legend
High resolution image and legend (81k)
We co-cultured bone marrow stromal cells (BMSCs) from R26R reporter mice with Cre+ multipotent progenitor cells isolated from postnatal brain and grown as neurospheres14. Previous studies have shown that these two cell types can fuse with embryonic stem cells in vitro6, 7. After 4 days in vitro (DIV), a small proportion of -gal+ cells were found in these co-cultures (1 to 2 cells per 80,000 cells) (Fig. 1d). Importantly, most -gal+ cells at 4 DIV had two or more nuclei, an observation that was confirmed by electron microscopy (Fig. 1e, i; see also Supplementary Fig. 1). This is consistent with the generation of -gal+ cells by fusion. Notably, after 10 or 15 DIV -gal+ cells formed small colonies and some of these cells were mitotic (Fig. 1g and inset). Cells in these colonies were invariably mononucleated, suggesting that with time and cell division the nuclei of these cells fuse or supernumerary nuclei are eliminated. In addition to neurosphere cells, R26R BMSCs were also co-cultured with primary cultures of Cre+ fibroblasts. Three independent co-cultures did not yield positive cells, suggesting that not all cell types are equally capable of fusion in culture. To further confirm that -gal expression was due to cell fusion, Cre+ neurosphere cells were labelled with 5-bromodeoxyuridine (BrdU) and then co-cultured for 5 days with R26R BMSCs. Most bi-nucleated -gal+ cells in these cultures had only one of the two nuclei labelled with BrdU (Fig. 1h–k), further confirming the reliability of this method to detect fusion events.
These results confirm previous work demonstrating that cell fusion occurs spontaneously in vitro6, 7. To study cell fusion in vivo, R26R reporter mice were lethally irradiated, and 2 days later were grafted with bone marrow from mice constitutively expressing Cre recombinase and green fluorescent protein (GFP) under the control of the -actin promoter (-actin-Cre–GFP mice, see 'allogeneic bone marrow transplantation' section of Methods). We analysed the grafted mice at 2 (n = 3) and 4 (n = 3) months after transplantation. These mice showed significant levels of haematopoietic reconstitution, which was measured by flow cytometry based on the frequency of GFP+ cells in peripheral blood (from 54.6% to 79.8% of nucleated blood cells). Brain, liver, heart, gut, kidney, lung and skeletal muscle from these mice were serially sectioned and stained for the presence of X-gal+ cells. In all animals, cells labelled with -gal were only found in brain, heart and liver, and not in the other organs studied (Table 1). As a negative control, we grafted R26R mice with bone marrow from wild-type mice. We did not find -gal+ cells in these animals, demonstrating that the reporter was not inappropriately activated, even after irradiation.
BMDCs fuse with hepatocytes in fumarylacetoacetate-hydrolase-deficient mice8, 9. Consistent with this finding, we also observed fused hepatocytes in our grafted mice (Fig. 2). -gal+ hepatocytes expressed albumin (a characteristic hepatocyte marker) (Fig. 2, h) but were negative for CD45 (a haematopoietic marker) (data not shown). Electron microscopy confirmed that these fused cells were typical hepatocytes with no features of haematopoietic cells (Fig. 2c). They contained glycogen granules, complete desmosomes and bile canaliculi (Fig. 2d). Two months after transplant most -gal+ hepatocytes co-expressed the donor marker GFP (Fig. 2e, f); however, a small fraction of fused hepatocytes were GFP negative. This fraction had increased 4 months after transplantation (Table 1). This result suggests that after fusion donor genes may be inactivated/eliminated over time.
Figure 2 Fusion of hepatocytes with BMDCs after bone marrow transplantation. Full legend
High resolution image and legend (119k)
At 2 and 4 months after transplantation, -gal+ cells were detected in the cerebellum, where labelled cells displayed the typical location and morphology of Purkinje cells (Fig. 3a). Two -gal+ cells were embedded in plastic and serially sectioned for light and electron microscopy. Serial reconstruction of the soma of these cells demonstrated the presence of two nuclei (Fig. 3b; see also Supplementary Fig. 2). Notably, the two nuclei presented very different morphologies: one had a wrinkled surface with multiple invaginations and a single nucleolus, typical of Purkinje cells15, whereas the second nucleus showed a uniform spherical shape with multiple nucleoli, suggesting a different origin (Fig. 3b, c). Electron microscopy analysis confirmed that these cells were Purkinje neurons (Fig. 3c) with typical Purkinje cell somata and organelle distribution, including structures with the characteristics of synaptic contacts (Fig. 3d). No signs of degeneration or abnormal structures were observed in the cytoplasm of these cells (Fig. 3c, d). -gal+ Purkinje cells stained positively for the Purkinje cell marker calbindin (data not shown). This is the first direct evidence, to our knowledge, showing that a neuron can fuse with a BMDC. These results suggest that previous observations of small numbers of Purkinje cells bearing markers of transplanted bone marrow cells5, 16, 17 may have arisen by fusion rather than transdifferentiation.
Figure 3 Purkinje cells fuse with BMDCs after bone marrow transplantation. Full legend
High resolution image and legend (74k)
The third organ where -gal+ cells were found was the heart (Fig. 4). The -gal+ cells were integrated into the myocardial wall and had a morphology and alignment that was indistinguishable from the surrounding cardiac muscle fibres (Fig. 4a–d). At the electron microscopy level, the fused cells had the morphology of mature cardiomyocytes, including developed filament bands and mature intercalated discs with desmosomes and GAP junctions connecting to neighbouring fibres (Fig. 4e). In addition, fused cardiomyocytes expressed cardiac troponin I (Fig. 4g). As seen in the liver, GFP was expressed in most of the fused cardiomyocytes at 2 months after transplantation (Table 1). In contrast, fused cardiomyocytes lost GFP expression at 4 months (Fig. 4h). It has been suggested that haematopoietic stem cells can partially restore the infarcted heart by transdifferentiation, giving rise to new myocardium4. Our results suggest that BMDCs can fuse with cells within the heart to form mature cardiomyocytes, but it remains unknown whether any new cardiomyocytes are generated as a result of this fusion.
Figure 4 BMDCs fuse with cells in the heart. Full legend
High resolution image and legend (62k)
Bone marrow cells include haematopoietic cells as well as mesenchymal cells and possibly other cell types. To test whether haematopoietic cells were participating in fusion events in vivo we used donor mice in which Cre recombinase was knocked into the CD45 locus (CD45-Cre). As CD45 is specifically expressed by haematopoietic cells18, 19, recombination should only occur in fusions involving cells of the haematopoietic lineage. To confirm the CD45-Cre expression pattern these mice were mated with R26R reporter mice. We observed widespread -gal expression in haematopoietic stem cells, bone marrow cells, and blood cells but not in other tissues such as brain, liver, or skeletal muscle (Supplementary Fig. 3). Nonetheless, we cannot rule out the possibility of CD45-Cre expression by very rare cells in other tissues or that non-haematopoietic cells might transiently activate CD45 expression during whatever nuclear reprogramming might occur after cell fusion.
CD45-Cre bone marrow cells were injected into four lethally irradiated R26R mice to look for fusion events (see 'Congenic bone marrow transplantation' section of Methods). The engrafted mice were analysed 10 months after transplantation. Consistent with the above observations, we found -gal+ hepatocytes in all four mice (Table 1; see also Supplementary Fig. 3e). -gal+ cardiomyocytes were found in two of the four mice, and -gal+ Purkinje cells were observed in one mouse (Table 1; see also Supplementary Fig. 3). As a negative control, we lethally irradiated seven R26R mice and transplanted them with 5 105 R26R bone marrow cells. No -gal+ cells were observed in these mice. These experiments suggest that haematopoietic cells fuse in vivo with cells in liver, heart and brain, but this does not rule out the possibility that other types of BMDCs might also participate in fusion.
In R26R mice that had been transplanted with bone marrow cells from -actin-Cre–GFP mice, we looked for evidence of transdifferentiation in the grafted animals. GFP+ cells that were negative for -gal (that is, they did not fuse with recipient R26R cells) exhibited characteristics of microglia in the brain, of Kupffer or pit cells in the liver, and of macrophages in the heart (Supplementary Fig. 4). Each of these cell types are of haematopoietic origin20-22 and can fuse under certain conditions23, making them candidates for the haematopoietic cells that fused with resident cells. In contrast, no GFP+/-gal- cells with the appearance of neural cells, hepatocytes, or cardiac muscle cells were observed. This suggests that cell fusion is the major mechanism by which haematopoietic cells can contribute to these tissues; however, our data do not rule out the possibility of rare transdifferentiation events, especially by other cell types or under other experimental conditions.
Our results suggest that BMDCs fuse with selective cell types in three organs. We did not observe evidence of fusion in skeletal muscle, gut, kidney, or lung in these experiments. The lack of evidence for fusion in these organs could be due to a lower efficiency of Cre-mediated recombination in these tissues or lower expression of the reporter gene (Supplementary Fig. 5). Alternatively, fusion may only occur in these tissues at a lower rate or under other experimental conditions, such as after injury.
With the exception of irradiation, the mice used in these experiments were healthy and did not have any pre-existing injury or pathology. Reconstitutions involving the -actin-Cre–GFP mice were allogeneic and therefore could have experienced graft-versus-host injury. However, reconstitutions involving CD45-Cre mice were congenic and did not involve any histoincompatibility. Although qualitatively similar results were observed in both contexts, considerable variation was observed from mouse to mouse in the extent to which fusion was observed. Therefore, the efficiency of somatic fusion in vivo is probably influenced by many variables.
Our results raise the fundamental question of whether fusion between haematopoietic cells and cells of the brain, liver and heart has a physiological role in the development or maintenance of these organs. Interestingly, many hepatocytes and cardiomyocytes under normal conditions have two or more nuclei22, 24. To our knowledge this is the first study to demonstrate Purkinje cells with two nuclei, but other studies have suggested that these neurons can be polyploid25, 26. Our results suggest that cell fusion may be the mechanism by which these cells become multinucleated or polyploid. Genetic material derived from blood cells may contribute through cell fusion to the survival and function of cells in different organs. Previous studies have shown that fused cells are positively selected during hepatic degeneration, helping to rescue a mutant mouse deficient for fumarylacetoacetate hydrolase8, 9. Our observation that fusion is a major mechanism by which BMDCs contribute to the heart, liver and brain draws into question the rationale for clinical procedures based on the idea that transdifferentiation of BMDCs can lead to the de novo generation of heart or brain cells. Additional studies in animal models will be required to determine whether fusion by BMDC cells can be used in reparative cell therapy.
Methods
Cell cultures Bone marrow cells from -actin-Cre or R26R transgenic mice were collected by flushing tibias and femurs with RPMI medium 1640 (Gibco BRL) supplemented with 3% fetal calf serum. Red blood cells were depleted using ice-cold ammonium chloride (140 mM in Tris 17 mM), and bone marrow cells were plated at a density of 2–4 107 cells per 9.5 cm2 in Iscove's modified Dulbecco's medium (IMDM; Gibco BRL) supplemented with 10% fetal calf serum, 100 U ml-1 penicillin, 100 mg ml-1 streptomycin and 10 mg ml-1 glutamine (complete IMDM medium). The non-adherent cell population was removed after 48 h and the adherent BMDC layer washed once with fresh medium; cells were then continuously cultured for 1–4 weeks.
For neurospheres, brain subventricular zone from 5–10-day-old -actin-Cre or R26R mice was collected. After papain dissociation, neurospheres were cultured and expanded in the presence of both epidermal growth factor (20 ng ml-1; Peprotech) and fibroblast growth factor-2 (10 ng ml-1; Peprotech), as described27. For BrdU labelling, neurospheres were cultured for 15 min in the presence of 2 µM BrdU, washed, and expanded for two additional passages before being cultured with BMSCs. This procedure labelled 70–80% of the neurosphere cells. Fibroblasts were cultured as described previously28.
Co-cultures BMDCs and dissociated neurospheres were mixed in a 1:1 ratio and plated on Matrigel-coated dishes (BD Bioscience) at a density of 2 105 cells ml-1 in complete IMDM medium. BMDCs and primary fibroblasts were cultured in a 1:1 ratio on plastic dishes in complete IMDM medium. After 4–15 days, co-cultures were washed and fixed in 2% paraformaldehyde for 10 min and analysed by X-gal staining or immunohistochemistry. As a negative control, R26R BMDC monocultures were grown for 15 days in the presence of conditioned medium or cell extracts from Cre-expressing cells (data not shown). Cell extracts from Cre-expressing cells were freshly prepared by two freeze/thaw series and added to the culture medium.
Animal care and bone marrow transplant Animal care and all procedures were approved by the Institutional Animal Care Committees at UCSF and the University of Michigan.
Allogeneic bone marrow transplantation Homozygous mice expressing Cre recombinase under the control of the hybrid regulatory element CMV enhancer -actin promoter11, and homozygous mice expressing GFP under the same promoter29 were bred to generate -actin-Cre–GFP mice for use as bone marrow donors. Bone marrow cells from 8–10-week-old Cre–GFP+ mice were extracted as described, and 10–20 106 cells were intraperitoneally injected into R26R mice irradiated with a single whole-body dose of 7.5 Gy. To avoid allograft rejection, mice that received the bone marrow transplantation procedure were treated one week before transplantation and 3 weeks after transplantation with Neoral cyclosporine (100 mg l-1; Novartis) in the drinking water. Drinking water was acidified and contained neomycin sulphate (1 mg l-1; Sigma) to suppress pathogens.
Congenic bone marrow transplantation In an independent experiment 8–10-week-old CD45-Cre 'knock-in' mice on a C57BL/Ka-Thy1.1 background were used as donors of bone marrow cells. The generation of CD45-Cre knock-in mice will be described elsewhere (E. Schaller and K.P., manuscript in preparation). Eight-week-old R26R mice on a C57BL/Ka-Thy1.2 background were used as recipients. Approximately 5 105 bone marrow cells were injected into the retro-orbital venous plexus of R26R mice lethally irradiated with two doses of 5.7 Gy each. The drinking water of the transplanted mice contained neomycin sulphate (1 g l-1) and polymyxin B sulphate (1 106 U l-1) to suppress pathogens. On a monthly basis after transplantation, mice were bled and the peripheral blood was stained with antibodies against Thy1.1 and haematopoietic markers to confirm reconstitution.
CD45-Cre knock-in mice were bred with R26R to obtain CD45-Cre/R26R mice. The -gal expression pattern in these mice was examined by X-gal staining of tissues and by fluorescein di--D-galactopyranoside (Molecular Probes) staining of haematopoietic cells. Bone marrow cells were incubated for 5 min at 37 °C with 10 mM fluorescein di--D-galactopyranoside in a hypotonic solution (1:1 staining medium (HBSS plus 2% calf serum):distilled water). Haematopoietic stem cells (Sca-1+ c-Kit+ Flk-2- lineage- cells)30 were analysed for -gal expression using a FACS Vantage flow-cytometer (Becton-Dickinson) (Supplementary Fig. 3a).
Tissue collection After 2, 4 or 10 months mice were anaesthetized and transcardially perfused with 0.9% saline followed by 50 ml 4% paraformaldehyde or 2% paraformaldehyde plus 0.25% glutaraldehyde. Brain, spinal cord, liver, lung, kidney, heart, skeletal muscle and gut were dissected. Brain and one liver lobe were serially cut in 50-µm vibrotome sections. The rest of the liver and other tissues were cryopreserved and frozen in optimum cutting temperature compound (Sakura-Finetec) at -80 °C. Serial 10- or 50-µm sections were cut in a cryostat and stained with X-gal or by immunohistochemistry.
X-gal staining and immunohistochemistry Specimens were placed in phosphate buffer containing 10 mM K3Fe(CN)6 and 10 mM K4Fe(CN)6 along with the -gal substrate X-gal (1 mg ml-1) (Molecular Probes) at 37 °C for 8–12 h. Antibodies against albumin (A-6684; 1:100) and calbindin (C-9848; 1:1,000) were from Sigma, cardiac troponin I (sc-1881; 1:1,000) was from Santa Cruz Biotechnology, BrdU (M0744, 1:100) was from DAKO, CD45 (558750, 1:100) was from BD PharMingen, and Iba1 was a gift from Y. Imai. Secondary antibodies anti-mouse-, goat- or rabbit-IgG (H + L) (Cy-2, 1:400; Cy-3, 1:400; biotinylated, 1:500) conjugated were from Jackson Immunoresearch.
Plastic embedding and electron microscopy Fifty-micrometre -gal-stained sections were post-fixed with 1% osmium and 7% glucose for 2 h, rinsed, dehydrated and embedded in araldite (Durcupan, Fluka). Semi-thin sections (1.5 µm) were cut with a diamond knife and stained lightly with 1% toluidine blue. Semi-thin sections were re-embedded in an araldite block and detached from the glass slide by repeated freezing (liquid nitrogen) and thawing. Ultra-thin (0.05 µm) sections were cut with a diamond knife, stained with lead citrate and examined under a Jeol100CX electron microscope.
Supplementary information accompanies this paper.
