Potential conflict of interest: Nothing to report.
A large portion of hepatocytes are polyploid cells, thought to arise through endoduplication followed by aborted cytokinesis. However, several recent reports describing liver cell fusion with exogenously derived bone marrow cells have been published. The exact significance of this finding is unclear, because the adopted protocols involve ablation regimens, damaged livers and artificial injections of adult cells. By creating chimeric mice bearing distinct reporter genes (LacZ and GFP), we show that in an unperturbed setting, hepatocytes carrying both markers can be detected via immunohistochemistry and polymerase chain reaction analysis. To further corroborate these findings with a direct visualization of the chromosome content at the single-cell level, we performed genotype analysis via fluorescence in situ hybridization on XY/XX chimeric mice with a Y chromosome–specific paint and an X chromosome–specific bacterial artificial chromosome clone probes. Conclusion: This technique confirmed the occurrence of cell fusion in adult mouse liver. (HEPATOLOGY 2008.)
Cell fusion is known to physiologically occur in specialized cells such as muscle, trophoblast, and osteoclast cells, but it is not usually thought to occur in the normal liver. Although a large portion of liver cells have a 4n, 8n, or even larger DNA content,1 this has been attributed to DNA duplication followed by aborted cytokinesis.2, 3 However, several recent reports in which exogenous bone marrow cells have been transplanted in recipient mice have shown that cells bearing donor-derived cellular markers can be found in host livers. This undisputable finding can be explained in at least two ways. According to some investigators, these results are due to adult stem cell plasticity, a recently recognized property that allows cells of a specific cell lineage (for example, a hematopoietic stem cell) to adopt functional abilities of a different lineage (for example, the liver).4, 5 Alternatively, cells bearing both donor and recipient cell markers could originate through cell fusion, a process in which gene expression from both contributing cells can coexist.6, 7 In this regard, the liver has been the most investigated tissue. Since the first report of the rescue of a degenerative liver disease by bone marrow transplantation,8 data supporting either interpretation (plasticity and cell fusion) have been published.5, 9–23 Interestingly, fusion events have also been detected in other organs, though at a lower level.24–26 The relative relevance of cell fusion versus plasticity has not yet been settled, but it could be emphasized that the two mechanisms are not mutually exclusive.
Most of these reports have been gathered from very artificial settings in which mice are severely treated with irradiation and/or other ablation regimens before exogenous cells are injected into recipient animals. In addition, this phenomenon seems to be enhanced by the coexistence of chronic liver damage, making it difficult to discern whether it has any physiological significance or is largely artificial.
To investigate whether fusion occurs in physiological conditions, we devised a simple but straightforward protocol based on the production of chimeric mice as originally designed by Mintz and colleagues.27–29 In several seminal papers, these investigators pioneered the use of chimeric mice in the study of gene expression and other cellular processes, including cell fusion. They concluded that fusion occurs in muscle but not in the liver, although at that time single-cell markers were not available, and they had to rely mainly on biochemical analysis of enzyme isoforms. It is noteworthy that their experiments were not repeated when single-cell markers became available, and the fact that cell fusion did not occur in the liver has since been taken for granted, although some limitations of the enzyme markers were recognized by Mintz.2
GFP and Rosa26 hemizygous transgenic mice were used as donors of morulae. GFP CD-1 transgenic mice, with the GFP gene controlled by the cytomegalovirus promoter/enhancer, were a kind gift of Dr. Masaru Okabe, Osaka University, Osaka, Japan,30 and Rosa26 mice in C57/Bl6 background31 were obtained from Jackson Laboratories. Mice were maintained in accordance with Italian Ministry of Health guidelines.
GFP/Rosa 26 chimeric mice were obtained by the aggregation technique. Morulae were recovered by flushing the oviducts of 2.5 days postcoitum female mice with M2 medium (Sigma). The zona pellucida was removed from 8-16 cell embryos with Tyrodes acid (Sigma). Embryos were then washed with 1× phosphate-buffered saline (PBS) and then with M2 medium. GFP morulae were selected via fluorescent microscopy, aggregated in pairs with morulae from heterozygous Rosa26 mice, and incubated at 37°C overnight. Blastocysts developed from morulae aggregation were reimplanted in utero in 2.5 days postcoitum pseudo-pregnant females to continue their development in vivo.
Fluorescence in situ hybridization (FISH) studies were performed on chimeras obtained via embryonic stem cell (ESC) injection in C57/BL6 blastocysts according to standard techniques.32 Injected blastocysts were reimplanted as described above.
Mice were anesthetized with Avertin and perfused intracardially with 10 mL of 4% paraformaldehyde. Organs were rapidly removed and fixed in 4% paraformaldehyde overnight at 4°C. The following tissues were processed for cell fusion analysis: liver, brain, cerebellum, skeletal muscle, kidney, spleen. Fixed organs were washed with 1× PBS four times for 10 minutes, dehydrated in 30% sucrose by overnight incubation, then quick-frozen in optimal cutting temperature compound (OCT, Bio-Optica). The samples were stored at −70°C and removed for cryosectioning as needed.
Single Hepatocyte Recovery.
To obtain fresh suspensions of adult hepatocytes, liver perfusion was performed as described.33 Briefly, mice were perfused through the portal vein with pre-perfusion solution to remove blood for 4 minutes and then with perfusion solution containing collagenase (IV Type, Sigma) for 10 minutes, preheated at 37°C. The liver was then removed and collected in a Petri dish. After washing twice with L-15 medium (Sigma) supplemented with antibiotics (100 U/mL penicillin and 100 μg/mL streptomycin), it was mechanically dissociated. The suspension was filtered (70 μm) and then centrifuged at 50g for 2 minutes. The pellet was washed again and resuspended in 25 mL of medium. Parenchymal cells were enriched by Percoll gradient centrifugation (50g for 5 minutes). For immuno-histochemical analysis, 105 cells were plated in chamber slides and allowed to adhere for 2 hours in William's E medium (GIBCO-BRL) supplemented with 15% fetal bovine serum (Sigma) and antibiotics (100 U/mL penicillin and 100 μg/mL streptomycin). After two washings with 1× PBS, cells were fixed with 4% paraformaldehyde at 4°C. For DNA FISH analysis, 2 × 105 hepatocytes were plated on square glass coverslips in 6-multi wells and incubated at 37°C overnight with the same medium (see above). Hepatocytes were then washed three times with 1× PBS, fixed in ice-cold methanol for 10 minutes, and stored at 4°C until their use.
FISH was performed using three different probes for the identification of the heterosome. For the X chromosome, we used a locus-specific probe and a paint probe; for the Y chromosome, we used a paint probe. The bacterial artificial chromosome (BAC) clone RP23-113K2, encompassing the distal region of the X chromosome, was obtained from the Children's Hospital, Oakland, CA. This probe was labeled via nick translation using biotin-16-dUTP (Boehringer Mannheim) and was detected with Alexa Fluor 647-conjugated streptavidin antibody (Invitrogen). Flow-sorted DNA for X and Y chromosomes (M.A. Ferguson-Smith, University of Cambridge, Cambridge, UK) were labeled via polymerase chain reaction (PCR) with Spectrum Orange-dUTP (Vysis) and Spectrum Aqua-dUTP (Perkin Elmer). Briefly, the square coverslips in which hepatocytes were plated after perfusion were incubated in denaturation solution (FA/SSC) at 90°C for 1 minute and 45 seconds and then dehydrated with serial ethanol washing steps (70 ice-cold, 90, 100% for 3 minutes each). Probes were denaturated in the hybridization solution (50% dextran sulfate/SSC) at 80°C for 5 minutes, put in ice for 2 minutes, and reannealed at 37°C for 1 hour. Probes were then applied onto the slides, covered with the square coverslips and incubated overnight at 37°C in a humidified chamber. After washing 3 times with 50% formamide/2× SSC for 5 minutes and three times with 1× SSC for 5 minutes, the coverslips were incubated at 37°C for 30 minutes with the blocking solution (3% bovine serum albumin). Thereafter, for the detection of the locus-specific probe, the coverslips were incubated with 120 μL of antibody solution in a humidified dark chamber at 37°C for at least 1 hour. After washing three times with 4× SSC/0.1% Tween 20 for 5 minutes, the nuclei were counterstained with 4,6-diamidino-2-phenylindole (DAPI) for 10 minutes and dehydrated with ethanol series. (Further details for FISH protocols can be found at http://topolino.aecom.yu.edu.)
Interphase cells were imaged with an Olympus BX61 microscope that had an UPlanSApo 100× N.A. 1.4 objective, an Hg arc lamp for excitation, and narrow band filters for all fluorescent emission and was equipped with a Cooke SensicamQE camera with IPLab for image acquisition. Images of interphase cells for each slide were acquired for the Spectrum Orange, Cy5, and Spectrum Aqua dyes. An IP lab script was generated to acquire images, a differential contrast image was acquired first to ensure that binucleated cells were sharing the same cytoplasm and were not the result of high-density cell plating. Multiple focal planes were acquired for each channel to ensure that signals on different focal planes were included: eight focal planes for chromosome painting and 13 different focal planes for locus-specific probes were acquired.
Cryosections of 10-μm thickness were cut at −20°C from embedded tissues using an HM 505 N MICROM microtome. Tissue sections and single cells plated on chamber slides were washed three times with 1× PBS for 10 minutes and then processed for the detection of LacZ markers by a specific antibody. Blocking was performed with a solution of 10% goat serum (Sigma) in 1× PBS. Samples were stained with rabbit anti–β-galactosidase (β-gal) primary antibody (Chemicon [1:500]) and goat anti-rabbit Cy3-conjugate secondary antibody (Jackson Laboratories [1:500]). Organ sections and single hepatocytes were counterstained with DAPI. The images were acquired with a confocal laser scanning microscope (Leica TCS SP2; Leica, Wetzlar, Germany) equipped with a 63×/1.4 NA oil immersion lens using 405-nm (DAPI), 488-nm (GFP), and 543-nm (Cy3) laser lines (Photodetectors: DAPI, 410-460; GFP, 505-530; Cy3, 560-640). Before acquiring multistaining images, the intensity of excitation wavelengths and the power of photodetectors were adjusted to avoid crosstalk and autofluorescence.
Single hepatocytes from GFP/Rosa26 chimeras, Rosa26, GFP, and double transgenic mice were analyzed via PCR. After serial dilution in the medium, single hepatocytes visualized under an inverted microscope were aspirated with a glass capillary pulled with a flaming brown micropipette puller (Model P-87, Setter Instrument Co.) and modified with a microforge (Bachofer) in order to obtain an inner diameter of 50-100 μm and thus minimize the possibility of sucking more than one cell.
Genomic DNAs were amplified using GenomiPhi DNA Amplification kit (GE Healthcare Life Science UK Limited) and purified by ethanol precipitation according to the manufacturer's instruction. The amplified DNA was used as a template for the two separate amplifications.
To amplify the GFP and LacZ transgene, we used the following primers:
CCTGAAGTTCATCTGCACCA and GGGTGTTCTGCTGGTAGTGG for GFP and CCGATATTATTTGCCCGATG and GCCCTGTAAACGGGGATACT for LacZ, which gave rise to a 431-bp and 206-bp band, respectively.
All the reactions were performed in 25 μL of final volume with 0.050 U Green Taq polymerase (Promega Corporation U.S.A.), 1.5 mM MgCl2, 200 μM dNTPs, 500 pM of each primer, and 3 μL of amplification products from the Genomiphi reactions.
The termocycling conditions used for amplification consisted of an initial denaturation step at 94°C for 4 minutes, followed by 35 cycles of denaturation at 94°C for 20 seconds, annealing at 60°C for 30 seconds, 72°C for 20 seconds, and a final elongation step of 2 minutes at 72°C.
Coexpression of GFP and β-gal Proteins in a Subset of Normal Liver Cells from Chimeric Mice.
Building on Mintz's approach, we reasoned that by aggregating morulae from two different strains of transgenic mice expressing either the GFP or the β-gal proteins, we could derive animals with two genetically distinct liver populations, each bearing a single marker; the appearance of a third population displaying both markers would be the result of cell fusion (Fig. 1A). Therefore, we compared our findings in these chimeric mice to those obtained in wild-type nontransgenic as well as in mice transgenic for either GFP (GFP single transgenic) or LacZ (Rosa26 single transgenic) and in mice bearing both transgenes (GFP/Rosa26 double transgenic).
Liver tissue is highly autofluorescent, and this can interfere with typical fluorescent labeling. To provide strict controls for the entire procedure, liver tissues and cells from wild-type nontransgenic mice were used to threshold the background. The same setup was used to acquire images from the other controls (GFP or Rosa26 single transgenic and GFP/Rosa26 double transgenic) and from the chimeras. Spontaneous GFP fluorescence at the level of single cells is both nuclear and cytoplasmic, while antibody-mediated fluorescence of Rosa26 mice is mainly detected at the membrane and cytoplasmic levels.
We obtained four live chimeric mice that were positive for both markers by morula aggregation. Chimeric mice 3 and 4 (8 and 10 months old, respectively) were analyzed; unfortunately, the low percentage of chimerism in their liver (<3% GFP+ cells) prevented further analysis. The other two mice were used for our analysis (Fig. 1A).
The results obtained from the analysis of the 7-month-old chimeric mouse 1 are shown in Fig. 2. Liver sections from wild-type nontransgenic mice displayed no markers, whereas most (but not all) liver cells from GFP or Rosa26 single transgenic mice were positive for GFP or β-gal, respectively, as expected. No cells from these single transgenic mice were positive for both markers, while both markers were easily detected in most cells from GFP/Rosa26 double transgenic mice (data not shown). The same analysis performed in chimeric mice identified three kinds of populations, one GFP, one β-gal, and one bearing both markers (Fig. 2). In light of the stringent controls used, we interpreted this third population as the result of cell fusion.
The specificity of our approach is confirmed by the results obtained in other organs. The skeletal muscle, known to be a syncitium formed by many cells, displays both markers in the chimeric mouse (Fig. 3A) and in the double transgenic mouse, yet neither GFP nor β-gal were found in single positive GFP and Rosa26 mice, respectively (data not shown). All muscle fibers in the chimeric animals were positive for both markers, because they are the product of the fusion of many cells of the two different genotypes, albeit in different percentages, giving rise to positive fibers with various intensities. No cell bearing both markers were seen in brain sections (Fig. 3B), with the exception of the cerebellum, where several Purkinje cells were double-positive (Fig. 3C). Finally, cell fusion was not detected in the spleen or kidney (data not shown).
In order to further confirm our findings in hepatocytes, livers from the 2-month-old chimeric mice 2 and controls (nontransgenic, GFP, Rosa26, or double transgenic GFP/Rosa26 mice) were perfused; part of the cells were plated, and a part was collected for PCR analysis (see below). The plated hepatocytes were examined for GFP and β-gal markers; the results are shown in Fig. 4. No marker was detected in nontransgenic animals (Fig. 4A, first lane), while only GFP and β-gal activities were detected in GFP and Rosa26 mice, respectively (Fig. 4A, lanes 2 and 3). Cells displaying both markers were found only in double transgenic control mice (Fig. 4A, lane 4) and in the chimeric mouse 2 (Fig. 4B). In the chimeric mouse, examination of the percentage of cells bearing a single marker revealed that they were approximately 40% GFP+ and 60% β-gal+. The percentage of GFP/β-gal double-positive cells in chimeric samples was high. Altogether, 164 out of 577 binucleated and 55 out of 274 mononucleated cells for a total of 219 out of 851 (26%) showed both markers. This suggests that both synkaryons (fused cells with a single nucleus) and heterokayons (fused cells with distinct nuclei) are formed in this process.34
We next investigated the possibility that transcripts or proteins derived from the GFP or LacZ reporter transgenes could be transferred from one cell to another. This could result in the presence of both markers in a single cell, without cell fusion. This cannot be excluded a priori, because it has been demonstrated that proteins present in the environment can be captured by the cell and wrongly interpreted as an endogenous product.35 Therefore, we performed careful PCR analysis on single hepatocytes mechanically isolated (see Materials and Methods) after perfusion of chimeric mouse 2. Analysis was performed on 152 single cells, and at least one amplification product (LacZ or GFP) was detected in 81 cells. Thirty-two of these cells were positive for the LacZ transgene only, and 38 were positive for the GFP transgene only; both transgenes were detected in 11 hepatocytes. This analysis showed that both GFP- and LacZ-specific genomic bands were amplified from 13% of individual liver cells from chimeric mice and from the majority of cells from double transgenic mice, but never from liver cells obtained from single transgenic animals (Fig. 4C). This confirms that at least a subset of liver cells does contain both markers, supporting the conclusion that they are the product of fusion between two different genomes.
A Subset of Binucleate Liver Cells from XY/XX Chimeric Mice Show the Y Chromosomes in Only One Nucleus.
These data prompted us to validate our findings by a completely independent approach that allows investigation of fusion events at the genomic level (Fig. 1B) in XY/XX chimeric mice. A portion of chimeric mice will be a mosaic of both XY and XX cells. In these XY/XX mice, liver cells will contain both XY and XX cells with a variable percentage. If endoduplication/aborted cytokinesis is the only mechanism leading to binucleated cells, both their nuclei will be either XY or XX (or multiples of them). On the contrary, if fusion occurs, a population of binucleated cells containing Y chromosomes in only one nucleus will be detected. Similarly, if mononucleated polyploid hepatocytes derive exclusively from binucleated cells progressing through an S phase, the formation of a bipolar spindle, segregation of all chromosomes, and cytokinesis at telophase (as described by Guidotti and coworkers36), they should only have a pair number of X chromosomes (if derived from the XX component of the chimeric mouse) or as many X chromosomes as Y ones (if derived from an XY cell). On the contrary, if a cell fusion event is involved in the formation of mononucleated polyploid hepatocytes, a variable number of sex chromosomes could be found, when the synkarion originated from the fusion of an XX cell with an XY one.
Therefore, we studied chimeric mice produced by injection of male ESCs from the 129sv strain into wild-type C57 blastocysts. Of the four chimeric animals analyzed, two were balanced XY/XX chimeras (about 30% XY and 70% XX), one was an XY/XY male mouse, and the last was an XX/XY animal with a very low contribution of XY cells in the liver.
The Y chromosome was visualized with a paint probe giving a specific, relatively large signal (covering its specific chromosome territory), as preliminarily confirmed in both metaphases spreads and interphase splenic cells from three wild-type XY mice. In these cells, only one signal per cell was obtained (data not shown). We then analyzed the liver of these wild-type XY mice, showing that binucleated cells always have at least one Y chromosome per nucleus (Fig. 5). However, when the two XY/XX chimeras (hereafter referred to as FISH1 and FISH2 mice) were analyzed, several binucleated cells with at least one Y signal in one nucleus and none in the other were unequivocally detected (Fig. 6A,B and Fig. 7). X chromosomes were visualized with a paint probe and/or a locus-specific BAC probe (see Materials and Methods). Because X chromosomes cover more territory than Y chromosomes, X locus-specific probes can be scored with more accuracy than the X chromosome paintings; therefore, X locus-specific probes were used for all the analyses, while the paint probe was used in selected experiments. The spots in each nucleus were counted and compared with the number of Y-specific signals. This allowed us to detect 11% and 16% of binucleated cells in which at least a Y chromosome signal was present only in one nucleus in FISH1 and FISH2 mice, respectively. This suggests that cell fusion in the liver is not a rare event.
This conclusion is strengthened by the analysis of mononucleated polyploid cells. A subset of mononucleated cells where Y chromosome signals were detected had Y chromosome counts not matching that of X chromosome signals (at least 5%), as detected by the X chromosome BAC-specific probe (Fig. 6C). This is at odds with the pattern that would be expected if only endoduplication of genomic material accounted for the polyploidy of these cells.
Our results differ from those presented by Willenbring et al.,15 who looked for hepatocyte–hepatocyte fusion but did not detect it. However, their experimental protocol was different from ours. They injected wild-type hepatocytes in LacZ+, Fah−/− mouse livers and subsequently performed serial transplantations in Fah−/− mice, but were unable to detect cells positive for both Fah and β-gal activities. This discrepancy could be explained by the age at which hepatocyte transplantation was performed, because in our mice the cells of both phenotypes were already present at birth. In addition, we have not yet investigated the developmental stage at which fusion starts, but it is likely that it does not occur before the third week of adult life, because polyploid cells are very rare before this date.3 Finally, we did not investigate whether hepatocytes also fuse with other cell types. For this reason, we could not exclude that other liver cell types may partner in the fusion process.
It should be noted that although we used a chimera strategy that is, to our knowledge, the less invasive approach for the analysis of cell fusion in vivo, it has been reported that embryonic stem cells could fuse in vitro with adult stem cells.37–38 Although there are no reports of cell fusion during whole embryo manipulation, we cannot dismiss the possibility that rare fusion events might occur during the experimental procedures used to generate our mice (morulae aggregation or ESC injection into blastocysts). In this case, cells carrying both markers could be expanded during development and could be detected in the liver through our analysis. Although we have not formally excluded this possibility, we think it is unlikely, because in this case cell fusion would be a widespread phenomenon within the generated mice and should be detected in several tissues.
In addition to hepatocytes, we detected fusion events in Purkinje cells; however, we did not pursue this finding further, because the study was designed to focus only on liver cells. Interestingly, liver and Purkinje cells are the only cells that have been reported to fuse with exogenously provided bone marrow cells.10, 14, 39 Moreover, the possibility of fusion in normal Purkinje cells has recently been suggested.40 This suggests that these rare fusion events exploit the intrinsic properties of liver and Purkinje cells to fuse with cells of the same histotype (homotypic fusion), but that fusion with cells of other histotypes (heterotypic fusion) is not a frequent event.
Our data support the fact that fusion is a physiological event in the liver and is one of the mechanisms contributing to hepatocyte polyploidy (although, in principle, endoduplication and cell fusion are not mutually exclusive). However, its role in liver economy is not addressed by this study and thus awaits further experimentation. Two main hypotheses have been raised to explain polyploidy and multinuclearity in hepatocytes. According to the first hypothesis, polyploidy and multinuclearity are a response to the continuous stress in detoxification that liver cells must bear. Alternatively, fusion in liver cells could occur at their terminal differentiation stage, indicative of cells already on the road to senescence. In this second case, therapeutic exploitation of this phenomenon could be more difficult,7 because it would target mainly senescent cells. Further experiments will shed light on these hypotheses, as well as on the role of fusion in neoplastic transformation of liver cells. The experimental approach used here will make it possible to finely dissect the role of various genes in this process by simply aggregating morulae from different genetic (knockout) backgrounds and investigating the effect on the appearance of cell fusion in liver or in other cells.
We thank Prof. R. Dulbecco for encouragement; Prof. M. Perrella and Drs. A. Villa, A. Frattini, and G. Merlo for useful discussion; and Juan Pablo Casado for technical assistance. We also thank Michael Cammer (Analytical Imaging Facility) and Dr. Winfred Edelmann (Gene Targeting Shared Resource) at the Albert Einstein College of Medicine Genome Imaging Facility. Finally, we thank Dr. Antonia Follenzi and Simone Merlin for useful suggestions.